INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Antimicrobial agents are one of the costliest categories of drug expenditure in hospitals. Investigations in various clinical practice settings have indicated that as much as 50% of antibiotic use is inappropriate (1, 2). Widespread use of antibiotics in intensive care units is a potential cause of the emergence of nosocomial infections caused by antibiotic-resistant gram-negative bacteria (3, 4). Ventilator-associated pneumonia (VAP) is a frequent cause of death from nosocomial infection in critically ill patients (5). The estimated prevalence of VAP ranges from 10 to 65%, with a high fatality rate, especially when it is caused by antibiotic-resistant bacteria (6, 7). The prior administration of antibiotics has become a recognized significant risk factor for superinfection caused by antibiotic-resistant microorganisms (8). This emergence of antibiotic-resistant microorganisms in critically ill patients represents a new challenge for intensive care physicians, and several antibiotic control strategies have been designed to prevent this problem (3, 9, 10).

In the light of bacterial ecology, which has been evaluated for many years in our medical intensive care unit (MICU), we decided to apply a program to improve antibiotic use, including a restricted use and a rotation of antibiotics in empiric and therapeutic uses for VAP treatment. The main goal of this prospective before-after study was to determine whether the incidence of VAP attributed to antibiotic-resistant gram-negative bacteria could be reduced by a scheduled change of antimicrobial agents.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population and Design

This study was performed in the MICU (16 beds) of a university teaching hospital (1,500 beds), between January 1995 and December 1998. During this period, all patients admitted to MICU were potentially eligible for this investigation. Patients were excluded if they were younger than 18 yr of age, and if their duration of MICU hospitalization was less than 48 h. Enrolled patients required mechanical ventilation for more than 48 h. The study was approved by the Bordeaux University School of Medicine Human Studies Committee. Informed consent was not required since all procedures were routine.

We performed a before-after study to test the hypothesis that a new program of antibiotic use could minimize the incidence of VAP and the emergence of antibiotic-resistant microorganisms. We chose the ventilator-associated pneumonia as the nosocomial parameter for evaluation because it was the most frequent nosocomial infection occurring in our MICU.

The before-period lasted 2 yr (January 1995 to December 1996). In this period, ceftazidime associated with ciprofloxacin was widely prescribed for the empiric and therapeutic treatment of VAP. This was the combination therapy most often prescribed. If the empiric therapy was effective, no change was made, i.e., de-escalation was not envisaged. The duration of antibiotic treatment was solely determined by the physician who managed the patients. Microbiologic data and incidence of nosocomial infections (VAP, bacteremia, urinary tract and catheter-related infections) were supervised in our unit for 1992 by a physician in our team.

Guidelines dictating which antimicrobials should be prescribed for empiric and therapeutic use were written in November 1996 and applied in December 1996. This new program was defined as follows.

1. Restriction of ceftazidime and ciprofloxacin in empiric and therapeutic uses.

2. Rotation of antibiotics without favoring any one antibiotic.

3. Each administration of antibiotics was determined by one of the two investigators (D.G., G.H.).

The usage of antibiotics was monitored on a monthly basis. Therefore, regular checks were made on prescription of one or another antibiotic. No preference was given to any particular antibiotic. The average rotation cycle was based on consumption supervised by the investigators. The results of monthly antibiotic intake were available at all times on the department computer. Prescription of a particular antibiotic was made only after consultation with one of the two investigators. The rotation outline was as follows.

In the case of late VAP (occurring  Day 7 of mechanical ventilation): (1) a betalactam and an aminoglycoside were prescribed empirically for the time needed to receive the bacteriologic results; (2) while waiting for the bacteriologic results from bronchoalveolar lavage (BAL), the choice of betalactam was cefepime or piperacillin-tazobactam or imipenem or ticarcillin-clavulanic acid; (3) over a period of 4 mo the choice of the betalactam was as follows: cefepime for the first month, then piperacillin-tazobactam for the second month, then imipenem for the third month, then ticarcillin-clavulanic acid for the fourth month; (4) each cycle could last more than 1 mo, according to the frequency of late VAP; (5) the choice of an aminoglycoside was made between tobramycin, amikacin, netilmycin, and isepamycin. In the case of isolated antibiotic-resistant bacteria, the empiric treatment was retained as treatment for VAP when possible. The antibiotic associated with betalactam was initially an aminoglycoside. In fact, we withdrew prescription of quinolones. The rotation cycle of aminoglycosides began with amikacin. From the second month, according to the total consumption of cefepime and the frequency of VAP in our MICU, we prescribed piperacillin-tazobactam instead of cefepime. Tobramycin was associated with piperacillin-tazobactam. From the third month, and according to the consumption of the antibiotic, imipenem was used instead of piperacillin-tazobactam. In this case, the associated antibiotic was netilmycin. From the fourth month, ticarcillin-clavulanic acid replaced imipenem. We decreased the use of ceftazidime as much as possible; (6) if isolated bacteria were not antibiotic-resistant, de-escalation was started as soon as the bacteriologic results were received; (7) as this de-escalation was possible only after reading the bacteriologic results, no choice was made in advance. The effective antibiotic was chosen according to the previous month's intake.

In the cast of early VAP (occurring < Day 7 of mechanical ventilation): (1) a betalactam was prescribed empirically while awaiting the bacteriologic results from the BAL; (2) the rotation of betalactam was on a monthly basis and considering closely the previous month's consumption; (3) over 4 mo the choice of betalactam was amoxicillin- clavulanic acid for 1 mo, followed by cefotaxim for 1 mo, then ceftriaxone for 1 mo, and then cefpirom for 1 mo. This treatment was validated on reception of the bacteriologic results. The antibiotic associated with betalactam was initially an aminoglycoside or fosfomycin. We withdrew prescription of quinolones as much as possible. This antibiotic rotation was supervised by the two investigators (DG and GH). With this cycle, the choice of an antibiotic was provided for in the treatment for VAP. Each treatment was validated by one investigator. We decreased the use of ceftazidime as much as possible.

4. The duration of treatment for microbiologically documented VAP was 15 d; we stopped aminoglycosides after 5 d of treatment.

5. Every morning, one of the two investigators met with the MICU patient care physicians team (three physicians) to insure optimal understanding and compliance with the protocol.

6. Once a week, a meeting was held with the nurses in order to identify the nosocomial problems of the unit.

7. Every month, one of the investigators evaluated the consumption of antibiotics. The studied parameter was: a day of antibiotic treatment, which is defined as the quantity (expressed in grams) of antibiotic consumed, divided by the defined daily dose of the antibiotic. The results of the global consumption of antibiotics in our MICU are expressed in days of antibiotic treatment.

8. Every trimester, all investigators met with the Laboratory of Bacteriology of our institution to evaluate the incidence of the emergence of antibiotic-resistant microorganisms.

We did not perform routine surveillance cultures to determine the colonization on MICU admission and during MICU hospitalization.

The after-period lasted 2 yr, from January 1997 to December 1998.

During the 4-yr period of the study, the hygiene protocols were noted and displayed in each room of the unit, without fail.

Definition of Clinical Suspicion of Ventilator-associated Pneumonia

The clinical criteria used to define the diagnosis of VAP were the same for the 4 yr of study. VAP was defined as any lower respiratory tract infection that developed after 2 d of mechanical ventilation (MV). Clinical suspicion of VAP was defined as a new, progressive, or persistent (> 24 h) infiltrate on the chest radiograph, with two or more of the following criteria: (1) fever > 38.3° C or hypothermia < 36° C; (2) purulent endotracheal aspirate; (3) leukocytes count > 10,000/ mm³ or < 4,000/mm³. Every patient suspected of having pneumonia underwent fiberoptic bronchoscopy to obtain samples by BAL.

Bronchoalveolar Lavage

Fiberoptic bronchoscopy was performed within 24 h after VAP was suspected. Patients were ventilated with 100% oxygen, sedated with midazolam, and paralyzed with vecuronium to avoid resistance to the ventilator during the procedure. Topical anesthetics were not used. Heart rate, systemic blood pressure, and arterial oxygen saturation measured by pulse oximetry were monitored throughout the BAL. The trachea was suctioned before introducing the bronchoscope through an adaptator, which allows the maintenance of mechanical ventilation. The bronchoscope was then wedge-positioned in a segmental bronchus corresponding to the radiologic infiltrate without suctioning to avoid contaminating the working channel. Three 50-ml aliquots of sterile normal saline were infused through the working channel of the bronchoscope. The fluid was then withdrawn by hand suction.

Microbiology and Definition of Ventilator-associated Pneumonia

The specimens were immediately sent to the laboratory. The aliquot of BAL fluid was cytocentrifuged, and the air-dried slides were stained using the Gram and May-Grunwald-Giemsa stains. By microscopic examination, the percentage of infected cells and the presence and type of extracellular and/or intracellular bacteria were determined. An infected cell was defined as a polymorphonuclear leukocyte or an alveolar macrophage containing at least one microorganism. Pure BAL fluid (0.1 ml) and 0.1 ml of a 100-fold BAL fluid dilution were plated on MacConkey agar, Chocolate agar, and 5% sheep blood Columbia colistin-nalidixic acid agar. After overnight incubation in 5% CO₂ at 37° C, the colonies were counted and identified conventionally. Colonies with distinct morphologies were enumerated separately and the results expressed as CFU/ml.

Microbiologically confirmed cases of VAP required the isolation of bacteria in significant quantities from BAL samples:  10⁴ CFU/ml.

Only the first episode of nosocomial ventilator-associated pneumonia was taken into account.

Definition of Microorganism Susceptibility

VAP was considered to be caused by potentially resistant gram-negative bacteria when Pseudomonas aeruginosa, Burkholderia cepacia, Acinetobacter baumanii, or Stenotrophomonas maltophilia yielded significant concentrations from BAL fluid.

Several antibiotics tested against each bacterium cultured at a significant level were considered for comparison during the study period: amoxicillin, amoxicillin-clavulanic acid, ticarcillin, ticarcillin-clavulanic acid, piperacillin-tazobactam, cefoxitine, cefotaxime, cefoperazone, ceftazidime, cefepime, cefpirome, imipenem, gentamycin, tobramycin, amikacin, and ciprofloxacin.

Microorganism susceptibilities were determined using the criteria established by the Comite National de l'Antibiothérapie, the official committee in France responsible for this classification. Results are expressed as percentages of susceptible bacteria.

Data Collection

Data collection at study entry included date and time of MICU admission, sex, diagnosis, results of chest radiographs, and severity of illness indexes: simplified acute physiologic score (SAPS II) (11), and the total number of failed organs according to the definitions established in the organ dysfunctions and/or infection (ODIN) model (12). Data collected and entered daily included date and time of endotracheal tube insertion and duration of mechanical ventilation.

Statistical Analysis

Continuous variables were compared using Student's t test or, when inappropriate, the Mann-Whitney U test was used. Chi-square statistics were used for categorical variables or, when not appropriate, Fisher exact test was used. Differences between groups were considered to be significant for variables yielding a p value  0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

There were 3,455 patients admitted to the MICU during the 4-yr period of the study: 892 and 845 patients were admitted to the MICU in 1995 and 1996, respectively, and 919 and 799 patients were admitted in 1997 and 1998, respectively.

Six hundred fourteen patients were excluded because of MICU stay < 48 h. Two thousand thirty-three patients required mechanical ventilation for more than 48 h.

The percentage of patients requiring mechanical ventilation was similar between the before-period and the after-period (Table 1); 506 and 498 patients required mechanical ventilation in 1995 and 1996, respectively (before-period), and 526 and 503 patients required mechanical ventilation in 1997 and 1998, respectively (after-period).

In patients who required mechanical ventilation for more than 48 h, the MICU mortality rate was 32 and 30.4% in the first and the second periods, respectively.

A total of 523 patients were suspected of having a ventilator-associated pneumonia. In 26 cases (15 in the before-period and 11 in the after-period), the clinical status of the patients ruled out bronchoscopy. These 26 patients were excluded. During the 4 yr of this study, 497 patients met the criteria for bronchoscopy and bronchoalveolar lavage (Table 1). The number of patients with clinically suspected VAP decreased significantly between the before-period (total number, 294: 144 in 1995 and 150 in 1996) and the after period (total number, 203: 116 in 1997 and 87 in 1998). This decrease was statistically more significantly between the before-period and 1998 (p < 0.01).

Diagnosis of VAP

A total of 392 patients had pneumonia diagnosed by BAL. The incidence rate of VAP for patients who received mechanical ventilation for more than 48 h was 22.1% in the before period and 15.7% in the after-period. Indeed, the number of patients with microbiologically documented VAP significantly decreased from 231 patients in the before-period (in 1995: n = 115; in 1996: n = 116) to 161 patients in the after-period (in 1997: n = 82; in 1998: n = 79).

Characteristics of patients with VAP

Clinical characteristics of enrolled patients are summarized in Table 2. Severity of illness indexes and indications for mechanical ventilation were comparable between the two periods. An antimicrobial therapy was present during the 15 d preceding the onset of VAP in 151 episodes (65%) in the before-period and in 112 episodes (69.5%) in the after-period.

During the 4 yr of this study, the mean duration of mechanical ventilation before the occurrence of VAP was 11.7 d (Table 2).

The mortality rate of patients with VAP in the before-period was not significantly different from the mortality rate of patients with VAP in the after period (Table 1).

Distribution of Causative Microorganisms

A total of 561 bacteria were cultured at a significant concentration by quantitative culture of BAL fluid: 332 and 229 bacteria were isolated in the before-period and the after-period, respectively. Microorganisms isolated from the 392 episodes of VAP are summarized in Table 3. The institution of a program of restricted antibiotic use did not decrease the incidence of polymicrobial VAP (Table 3). During the before-period, the most frequently isolated bacteria were ENTEROBACTERIACEAE (n = 82, 24.7%), Staphylococcus aureus (n = 67, 20.2%), and Pseudomonas aeruginosa (n = 62, 18.7%). During the after-period, the most frequently isolated bacteria were Staphylococcus aureus (n = 54, 23.6%), ENTEROBACTERIACEAE (n = 52, 22.7%), and Pseudomonas aeruginosa (n = 47, 20.5%).

There were 140 (42%) and 79 (34%) potentially resistant gram-negative bacteria isolated during the before-period and the after-period, respectively (p = 0.06). There was no difference in incidence of P. aeruginosa between the two periods of the study (18.7 versus 20.5%). The occurrence rates of B. cepacia, S. maltophilia, and A. baumanii decreased between the two periods, but not significantly (Table 3).

The incidence of other gram-negative bacteria did not decrease during the study. Approximately the same incidence of ENTEROBACTERIACEAE responsible for VAP was observed between the two periods, with a decrease in the after-period, except for Proteus sp., which significantly increased from six bacteria (2.5%) to 11 bacteria (7.2%) (p = 0.02).

Staphylococcus aureus and other gram-positive cocci were responsible for VAP in an insignificantly different proportion in the after-period.

As we indicate in Table 2, VAP occurring before Day 7 of mechanical ventilation were more frequent during the first period (31/231: 13%) than during the second period (15/161: 9%). During the first period, potentially resistant bacteria were responsible for early VAP. Bacteria responsible for early VAP were more sensitive to antibiotics in the second period.

Bacteria Susceptibilities

The susceptibilities to antibiotics of potentially resistant gram-negative bacilli increased after the set up of an antibiotics policy program. In fact, for these bacilli, the percentage of susceptibility to most of the antibiotics increased from the before-period to the after-period. Analysis of the susceptibilities of P. aeruginosa, B. cepacia, S. maltophilia, and A. baumanii is detailed in Table 4. Ciprofloxacin in particular showed an improved efficacy.

No increase in the susceptibility of the other gram-negative bacteria was found between the two periods. Nevertheless, a high percentage of these bacteria were sensitive to each antibiotic regimen.

Tests also revealed beneficial results with regard to the S. aureus responsible for VAP. Indeed, the percentage of methicillin-sensitive S. aureus increased from 40% in the before-period to 63% in the after-period (p < 0.05). The activity of fluoroquinolones against S. aureus significantly increased from 40 to 68.5% (p < 0.05), with better results in 1998: susceptibility to ofloxacin increased from 40% in the before-period to 73% in 1998 (p < 0.01).

The accuracy of antibiotic therapy was different between the two periods. A greater proportion of patients received an inadequate therapy in the before-period (73/231 cases of ventilator-associated pneumonia: 32%) compared with the after-period (32/161 cases of VAP: 20%) (p < 0.05).

Antibiotic Regimen Uses

The antibiotic control policy led to a reduction in total antibiotic usage. This holds true for most of the antibiotics. The results of the total consumption of antibiotics are detailed in Table 5. Concerning the total consumption of ciprofloxacin and ceftazidime during the two study periods: the total number of ciprofloxacin days decreased from 8,022 in the first period to 1,252 in the second period (difference of 84%). The total number of ceftazidime days decreased from 993 in the first period of 670 in the second period. The consumption of ofloxacin significantly decreased between the two periods of this study, respectively by 88%.

Among the betalactams, intake decreased except for piperacillin-tazobactam and cefepime. Cefepime was available in our institution during the second semester 1996. Piperacillin-tazobactam was used with a higher frequency in the second period. Cefepime and piperacillin-tazobactam showed a faster turnover in cases of VAP caused by potentially resistant gram-negative bacilli.

Among the class of aminoglycosides, the antibiotics turnover comprised amikacin, tobramycin, netilmycin, and isepamycin. In our institution, isepamycin has taken the place of amikacin since the beginning of 1998. The overall intake of aminoglycoside fell, except for isepamycin, introduced into our institution during the second period only.

Usage of vancomycin decreased in parallel with a fall in the occurrence of methicillin-resistant Staphylococcus.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Regarding the bacterial ecology and the increasing antimicrobial resistance in our MICU during 1995 and 1996, we attempted to achieve better use of antimicrobial agents with their restricted use and their rotation. This new objective was adopted at the end of 1996. We undertook a prospective before-after study on a series of consecutive MICU patients who required mechanical ventilation for > 48 h, using strict criteria to define VAP. Our protocol was listed as an important goal of quality assurance program. We wanted to determine whether an improved use of antibiotics through specific guidelines, supervised by two physicians of our team, could minimize the antibiotic resistance of gram-negative bacteria responsible for VAP. Our results indicate that the incidence of clinically suspected VAP and microbiologically documented VAP decreased significantly after the scheduled change of antibiotic classes. We also found a trend of a lower incidence of potentially antibiotic-resistant gram-negative bacilli. The susceptibilities of potentially antibiotic-resistant bacteria to antibiotic regimens significantly increased, especially for P. aeruginosa and B. cepacia. Similarly, the incidence of methicillin-resistant S. aureus responsible for VAP decreased significantly. The accuracy of antibiotic therapy was better in the second period. Nevertheless, we did not find a significant reduction of MICU mortality rate either in all patients who required mechanical ventilation for more than 48 h or in patients with VAP between the two periods.

The attributable mortality of VAP in critically ill patients remains controversial. In a recent and rigorous study, Heyland and coworkers (13) reported that the time of onset of VAP, the type of organism ("low" and high risk organisms), and the appropriateness of early empiric antibiotic therapy, did not influence the risk of mortality. The initial objective with this prospective before-after study was to determine whether the incidence of VAP attributed to antibiotic-resistant gram-negative bacteria could be reduced by a scheduled change of antimicrobial agents. An accurate explanation of the same MICU mortality rate in patients with VAP between the two periods would need a different prospective-matched cohort study, with a matching procedure between patients with VAP and patients who did not develop clinically suspected VAP. It was not the aim of this initial study.

A greater proportion of patients received an inadequate therapy in the before-period. This result could be associated with the better ecology in the second period, especially for early VAP occurring before Day 7 of mechanical ventilation, and with the better susceptibilities of potentially antibiotic-resistant bacilli during the second period.

A new program of antibiotics administration led to a decrease in the occurrence of VAP. This decrease especially concerned early VAP: pneumonia occurring before Day 7 of mechanical ventilation was more frequent during the first period than during the second period. Therefore, we found an overall reduction in the incidence of infection with antibiotic-resistant organisms. This issue seems paradoxal: in general, early VAP is not due to resistant pathogens. Our ecology was so worrying during the first period that in our MICU, potentially resistant bacteria were responsible for early VAP. Bacteria responsible for early VAP were more sensitive to antibiotics in the second period. The incidence of late VAP did not differ between the two periods. Because the severity of certain cases necessitates longer periods of mechanical ventilation, and because of the immunodepression which accompanies organ dysfunction occurring in the ICU, we wonder whether it is really possible to avoid late VAP, with potentially resistant organisms, and whether it is just possible to have an effect on the antimicrobial resistance.

Antimicrobial resistance seems to be related to antimicrobial use. Several reports have suggested this relation (14). McGowan and Gerding (16) suggested that (1) increasing duration of patient exposure to antimicrobial agents increases the likelihood of colonization with antibiotic-resistant microorganisms; (2) the changes in antimicrobial agent uses coincide with the changes in prevalence of resistance; (3) antimicrobial resistance is more prevalent in nosocomial bacterial strains than those in community-acquired infection; and (4) intensive care units have the highest rates of antimicrobial resistance, but they also have the highest rates of antibiotics prescription.

Bacterial ecology with the emergence of resistant-antibiotic organisms was of great concern to us in 1996. During this period, more than 40% of VAP were caused by potentially antibiotic-resistance gram-negative bacilli. Similarly, during this period, 60% of S. aureus responsible for VAP were methicillin-resistant. This bacterial ecology was worrying. Without making ourselves unpopular with the other physicians in our team, a program to control and restrict antibiotic use had to be set up. This antimicrobial stewardship included limiting the use of ceftazidime and ciprofloxacin, because of a previous widespread overuse and a possible "de-escalation" in antibiotic choice, with an appropriate control of dosing and duration of treatment.

The effectiveness of antimicrobial control has already been reported (14, 16). With the control of gentamycin resistance through the restriction of gentamycin and replacement with amikacin, Gerding and coworkers (17) showed significant reductions in gentamicin resistance during this restriction period. Meyer and coworkers (7) showed that restricting the use of ceftazidime allowed them to control an outbreak of ceftazidime-resistant Klebsiella pneumoniae. Pestotnik and coworkers (18) used a computer-assisted decision support system to decrease the total pharmacy acquisition costs of antibiotics and stabilize antimicrobial resistance patterns.

We chose to compare the 2 yr before with the 2 yr after the institution of a new program of antibiotics use policy. We did not know the timespan necessary to improve the bacterial ecology in our MICU. We decided to report this evaluation at the end of 1998, because of the good results obtained. Our choice of parameter for evaluation was ventilator-associated pneumonia, because of the lower incidence of nosocomial bacteremia and urinary tract and catheter-related infections. These first results were encouraging. We observed a significant reduction in incidence of clinically suspected VAP and microbiologically documented VAP. This result has been also reported by Kollef and coworkers (3). As suggested by Niederman (19) in a recent editorial, antibiotics could play a role in the pathogenesis of VAP, in predisposing infection with resistant organisms.

The decrease in the incidence of potentially antibiotic-resistant bacteria was not statistically significant. This could be explained by the low total number of microorganisms, and also by better results in 1998 than in 1997, during the after-period. Our results were encouraging for the decease in incidences of ciprofloxacin-aminoglycoside-third generation cephalosporin-resistant P. aeruginosa. The incidence of P. aeru-ginosa did not decrease, but its susceptibility to antibiotic regimen was greatly improved. The poor susceptibilities of P. aeruginosa to antimicrobial agents, especially ceftazidime and ciprofloxacin, that we had in 1995 and 1996 were a crucial factor for us in the decision to change antibiotic treatment.

Because quinolones penetrate extremely well into respiratory secretions, we often used ciprofloxacin as empiric and therapeutic treatment of VAP during the before-period. Its supervised restricted use allowed us to significantly improve the susceptibility to this antibiotic of P. aeruginosa, B. cepacia, and A. baumanii. While still on a program of rotation of antimicrobial agents, we decided to introduce ciprofloxacin in the empiric and therapeutic treatment of VAP again, avoiding the monotherapy. Its beneficial effect in the treatment of pneumonia is highly important because of the excellent pulmonary concentration (20).

An induction of chromosomal beta-lactamases in gram-negative bacilli by a repeated exposure to the same antibiotic could explain the cephalosporin resistance. The appearance of beta-lactamases has been identified as a consequence of the use of third-generation cephalosporins. This cephalosporin resistance could decrease with restricted use (21). The rotation of antibiotic use for empiric and therapeutic treatment of VAP could have restored a higher activity of cephalosporin. Our results confirm this hypothesis with a higher activity of ceftazidime, cefepime, and cefpirome for the majority of gram-negative bacteria. The replacement used during this rotation was a beta-lactam/beta-lactamase inhibitor combination such as cephalosporins, ticarcillin-clavulanic acid, or piperacillin-tazobactam, or a carbapenem. It is interesting to note that a real rotation of antibiotics could prevent new resistance. Indeed, the percentage of susceptibilities to the prescribed beta-lactams were equivalent between the two periods, with better results in the after-period. A more accurate antibiotics therapy allowed us to decrease the use of less effective antibiotics during the second period. Antibiotic therapy was easier to adapt during the second period in part because the bacterial ecology was better during this period. There would appear to be a link between an efficient use of antibiotics, bacterial ecology, and an accurately adapted empirical therapy.

Similarly, we improved the susceptibility of S. aureus responsible for VAP, without reducing its incidence. Cyclic use was more problematic for gram-positive cocci such as staphylococci, for which the number of effective agents is limited. In these cases, we supervised the restriction of quinolones, the rotation of antibiotics such as fosfomycin, rifampin, pristiniamycin, and the "de-escalation" if possible, with a tendency to reduce the consumption of vancomycin.

Even though the bacterial ecology studied in this article was specific to our own ICU, certain general recommendations may be used in all units. Without claiming to be a "recipe" or to be a rigid standardization, certain recommendations are universal. Faced with our ecology concerns, we "rolled up our sleeves," so to speak, and applied a few elementary rules (22, 23). An antibiotic therapy must be used following a regularly monitored ecology. The dosage and the rhythm of antibiotic administration must be taken into account. The initial choice of the antibiotics is important. Antibiotic rotation seems to prevent the emergence of resistance. The reevaluation on Day 3 of all treatment with antibiotics allows us to check the appropriateness of the initial therapy, to modify the therapy using antibiotics that are just as effective but which have less influence on the microbial ecology, to check the posology and the rhythm of administration, and to specify the total duration of treatment and the chosen association. A reevaluation on Day 10 of treatment allows us to control the length of treatment for microbiologically documented pneumonia. With regular readouts from the pharmacy on the prescribed daily dose and/or the days of antibiotic treatment, the monitoring of the antibiotics used could represent an indicator of nosocomial infection. The evolution in antibiotic resistance should be available, based on the results from the bacteriologic units. Efficient cooperation between departments (ICU, Bacteriology, Pharmacy) seems indispensible.

Our study has several limitations. First, this study was carried out in only one center, with a specific ecology. Second, we believe that a large study is needed to evaluate the impact of a new antibiotic use protocol on the ICU and hospital mortality caused by potentially antibiotic-resistant bacteria. A matched cohort study comparing patients with and without VAP during the same two periods could evaluate the attributable mortality of VAP, and the potential role of the improvement of bacterial ecology. Third, in this study, supervised rotation of antibiotics was one new strategy among others aimed at improving antibiotics usage, as well as restricted use of two antibiotics, and new rules for prescription of other antibiotics. We could not attribute the improvement of bacterial ecology to any particular new intervention. It would be difficult to define the responsibility of the rotation of antibiotics. Last, in our MICU, we did not perform routine surveillance cultures in order to evaluate the colonization with antibiotic-resistant bacteria. This study could not estimate the impact of the protocol on colonization in our unit, which was perhaps the intermediate phase between infection and VAP. In this study, we did not perform an analysis of the cost of antibiotics prescription. We only performed an analysis, between two periods, of the difference of total daily doses of some antimicrobial agents. A real and efficient turnover of antibiotic use necessitates the prescription of expensive antibiotics. We have chosen to wait before comparing the cost of our new policy. We are sure that in reducing the incidence of VAP caused by antibiotic-resistant microorganisms, a potentially more expensive antibiotic protocol could in fact be an economical maneuver in middle to long term. Further studies are needed to validate our results, and to demonstrate the importance of such protocol in order to minimize the emergence of resistance patterns.