Ventilator-associated pneumonia—Understanding epidemiology and pathogenesis to guide prevention and empiric therapy

INTRODUCTION

Nosocomial pneumonia is the second most frequent hospital-acquired infection and is associated with high morbidity, mortality and increased hospital cost 1. Nosocomial pneumonia occurs in 0.6-1% of hospital admissions and 7-44% of ventilated patients in the intensive care unit (ICU). Overall, it accounts for 13-18% of all nosocomial infections, prolongs hospital stay by 7-10 days and is associated with a crude mortality of 20-60% and an attributable mortality (the percentage of deaths which would not have occurred in the absence of pneumonia) of approximately 30%. In the ICU, pneumonia accounts for a higher proportion of nosocomial infections than in the hospital as a whole (35-45% in ventilated patients). Although the crude mortality is higher (40-80%) in the ICU, the attributable mortality in ventilated patients (27%) is similar to that seen in the hospital overall 2. Nosocomial pneumonia is independently associated with death in ICU patients (Table 1), although some recent studies have produced conflicting results 3,4.

In one of these studies, the predictors of mortality were examined in two populations of patients in the ICU, namely, those who survived and those who died. Nosocomial pneumonia was shown to be a significant factor associated with increased mortality, with an odds ratio of 2.08, similar to that of nosocomial bacteremia 3.

If the mortality rate in ventilated patients is examined in more detail, it can be seen that different mortality rates may be attributable to different causative organisms. For example, Pseudomonas spp. and Acinetobacter spp. may be responsible for a mortality rate of up to 43% and even > 70% 2. An effective strategy for the prevention of nosocomial infections in the ICU should significantly reduce mortality, but to assess potential strategies, it is essential to achieve a consistent diagnosis.

DIAGNOSIS

All studies regarding epidemiology, pathophysiology, risk factors, treatment, prognosis and prevention of pneumonia are hampered by the same problem — the difficulty in diagnosing pneumonia. Criteria used clinically to define nosocomial pneumonia include fever, new or progressive pulmonary infiltrate on X-rays, leukocytosis, purulent bronchial secretions and a tracheal aspirate Gram's stain showing > 25 leukocytes and < 10 squamous epithelial cells (a sign of sample contamination) per low-power field, associated with recovery of a potential pathogen 5. Fever and infiltrate, however, are by no means specific for pneumonia. In the case of mechanically ventilated patients in particular, there are many other alternatives which need to be considered. Fibroproliferation of late adult respiratory distress syndrome (ARDS), pulmonary embolism atelectasis, drug reactions, congestive heart failure and pulmonary hemorrhage all show similar clinical signs and may mimic pneumonia. The presence of microorganisms in a tracheobronchial specimen may represent colonization or infection. Thus, efforts have been made to develop appropriate diagnostic testing, including changes in plasma concentration of C-reactive protein 6 or changes in serum concentration of procalcitonin 7 to determine whether the patient has pneumonia, to identify the pathogen and to help define the best therapeutic regimen. Lung specimens taken for histology and culture are considered to be the ‘gold standard’ in the diagnosis of pneumonia. The criteria for histopathologic diagnosis of pneumonia include the presence of neutrophilic infiltration of the terminal bronchioles and of the alveoli, which may be filled with neutrophils, fibrinous exudates and cellular debris (resulting from lung necrosis) which may be seen by Gram's stain 8. For certain microorganisms, specific methods should be used. The etiological diagnosis is established by identification of the pathogen(s), generally by culture. There are major problems, however, with the gold standard method. Firstly, the invasive nature of the technique means that these specimens cannot be obtained for most episodes of pneumonia; secondly, if there is a lung biopsy or autopsy, it may not be obtained at the same time as the pneumonic episode, making it difficult to be sure that the findings relate to a particular clinical episode of pneumonia. Moreover, the empiric use of antibiotics to treat the episode may invalidate direct examination or culture results. In ventilated patients lung histology is usually only used in post-mortem studies.

Given this problem of diagnosis, improved, less invasive methods have been developed. The most important of these are listed in Table 2 [4,9–32]. It should be borne in mind that sensitivity and specificity are always measured in relation to another method which will, in itself, have advantages and disadvantages. Thresholds have evolved from various studies, for which the sensitivity and specificity vary, depending on how the criteria were balanced. Results from a recent meta-analysis suggest that the use of a pre-determined threshold concentration of bacteria for either protected specimen brushing (PSB) or bronchoalveolar lavage (BAL) may not be appropriate in all clinical settings and that incorporating the clinical likelihood of pneumonia into the evaluation could significantly improve the sensitivity and specificity of these tests 33. It can be seen from Table 2 that an indisputable diagnosis cannot be reached using any of the methods. Nevertheless, it is generally accepted that PSB and BAL are currently the most accurate methods for diagnosing pneumonia. Although the sensitivity and specificity of these methods are highly variable, a recent study has shown sensitivities of 82% and 91% and specificities of 89% and 78% for PSB and BAL, respectively, compared with histologic and microbiologic post-mortem lung specimens, in the diagnosis of pneumonia 34. However, other postmortem studies have found lower sensitivities and/or specificities 4,35,36.

Both of these methods can be performed either through a bronchoscope or blindly, using an endobronchial catheter. PSB involves the passage of a catheter into the distal airways and the extrusion of a small brush (protected inside the catheter by a biodegradable plug) for the recovery of secretions. BAL requires sequential instillation and aspiration of a physiological solution into a lung subsegment through a catheter or a bronchoscope wedged in an airway. Mini-BAL (mBAL) refers to BAL using a smaller volume of solution (10-20 mL, rather than 100-200 mL) instilled distally.

The specimen volume retrieved with PSB is much smaller (0.01-0.001 mL) than with BAL and the area of lung sampled is also smaller, such that the threshold for quantitative samples is usually higher for BAL (> 10⁴CFU/mL) than for PSB (> 10³CFU/mL). However, the risk of contamination is less for PSB and mBAL (protected) than for BAL. One advantage of BAL, however, is that it allows the sampling of enough material for direct examination by Gram's stain and some studies have shown that the presence of visible intracellular bacteria is a good predictor of ventilator-associated pneumonia (VAP) 37. In addition, direct examination may help in the choice of initial antibiotic treatment.

PSB and BAL have their own problems. For example, when an examination was repeated using PSB on the same patient twice, although 100% qualitative reproducibility was recorded, a 24-25% quantitative discordance in the numbers of microorganisms recovered was reported between the first and second examinations 16,38. This may have been caused by peripheral wedging of the brush, but means that for PSB, no quantitative threshold exists which absolutely separates patients with and without pneumonia 25. Also, many patients received antibiotics and it has been shown by Meduri et al. 39 that this can reduce the number of true positive results obtained by PSB quantitative culture from 77% in untreated patients, to 60% in those on antibiotic therapy. Recent studies have shown that quantitative analysis of endotracheal aspirates appears to be of reasonable diagnostic value 40,41. Investigations should be performed before introducing or changing antibiotic therapy 42. If possible, antibiotic therapy should be discontinued for 48 h before PSB or BAL are used 42. Increased levels of endotoxin has been shown to predict the prescence of Gram-negative pneumonia 43,44. New diagnostic methods, such as the measure of procalcitonin levels in bronchial secretions are being investigated clinically 7.

In summary, despite the improvement in diagnostic methods, not all patients are correctly diagnosed as having pneumonia, with both under- and over-diagnosis being a problem. Better communication is necessary to achieve optimal diagnosis.

ETIOLOGY AND PATHOPHYSIOLOGY

Table 3 shows the causative organisms of pneumonia in ventilated patients. The spectrum varies according to regional differences, underlying disease and previous antibiotic treatment. In many cases knowledge of the sources of these organisms helps in understanding the pathophysiology of the infection. Often, the source is endogenous, but the organism may also come from the environment, from other patients or from hospital staff (Figure 1). Devices and other equipment are currently less of a problem, now that they are recognized sources of infection and aseptic handling and equipment have improved. Cross-infection from other patients and staff may be minimized, provided that human resources are adequate and that motivation is high.

Episodes occurring within the first few days after intubation (early-onset pneumonia: 4 days has been proposed as an arbitrary time point) 45 are mostly caused by bacteria of the normal oropharyngeal flora. If certain factors are present, such as prior antibiotic therapy, prolonged hospitalization or special host factors, there may be a different spectrum of microorganisms with different antibiotic susceptibilities, and this has important diagnostic and therapeutic implications. Episodes occurring after 4 days of intubation (late-onset pneumonia) are mostly due to Gram-negative bacilli or S. aureus.

PREVENTION

Risk factors and associated preventive measures are summarized in Table 5. Some risk factors, such as age (> 60 years), pulmonary disease, organ failure, coma, gross aspiration, cigarette smoking, diabetes, hypotension, alcoholism, diseases of the central nervous system, chronic obstructive lung disease, azotemia and respiratory failure are patient-related 1 and thus difficult to modify. Other risk factors are dependent on drug prescription. The stress-ulcer prophylaxis regimen has been mentioned, but the use of certain other drags are also risk factors. Sedatives, corticosteroids and cytotoxic agents may impair host defences. Antibiotic prophylaxis or treatment given prior to an episode of nosocomial pneumonia may select for resistance and thus reduce therapeutic options. Endotracheal prophylactic gentamicin was shown to be ineffective in preventing pneumonia and associated with the occurrence of gentamicin-resistant bacteria 83.

Effective preventive measures have been demonstrated in several studies 84. Two studies have suggested that removal of subglottal secretions, using either a specifically designed endotracheal tube or a catheter, may decrease the risk of pneumonia 64,85. In the study by Valles et al., patients with continuous subglottic aspiration were less likely to develop pneumonia than those without subglottic aspiration 85. The incidence rate of VAP was 19.9 episodes/1000 ventilator days in patients undergoing continuous aspiration and 39.6 episodes/1000 ventilator days in the control group. The difference was caused by a significant reduction in the number of infections due to Gram-positive cocci and H. influenzae in those receiving continuous aspiration (p < 0.03). New tubes are currently being introduced which allow aspiration on a continuous basis, although these tubes do have problems in their own right, such as contamination and blockage of the lumen.

Another recommendation is that both endotracheal and gastric tubes are inserted orally, rather than nasally. Rouby et al. 86 showed the incidence of sinusitis to be linked to nasal placement. Moreover, nosocomial pneumonia occurred significantly more frequently in patients with maxillary sinusitis (67%) than in those without (43%) (p = 0.002). These results have been confirmed in another study 87.

A simple measure to partially prevent esophageal reflux is to put the patient in a semi-recumbent position, which has been shown to decrease the recovery rate of identical microorganisms in the stomach, pharynx and trachea (32%), compared with patients in the supine position (68%). However, this is not feasible for all ventilated patients 88.

The incidence of pneumonia is increased if the ventilator circuit is manipulated frequently and it is now recommended to change it only every 72 hours or longer. Some recent studies suggest that it can be safely maintained for up to 7 days 89. Condensate in the tubing can also be a reservoir of pathogens and should be drained away from the patient; heat and moisture exchangers might avoid the problem of condensation and prevent colonization, thus permitting the use of the same circuit for a longer period of time.

Many pathogens, such as P. aeruginosa and S. aureus, survive well in the external environment. Handling of devices and equipment should therefore be minimal. Aseptic technique and good hygiene, with frequent hand washing or gloves for staff, isolation of patients with resistant pathogens and appropriate disinfection of equipment can decrease environmental and cross-contamination.

Malnutrition is a known risk factor for nosocomial pneumonia and has led to the use of nutritional support in hospitalized patients. There is no clear evidence that nutritional support reduces the risk of pneumonia, however, and care must be taken regarding preparation, handling, route and volume of feeding. Enteral nutrition may stimulate the intestinal mucosa and the immune system 90, but any increase in gastric pH and volume could increase the number of potential padiogens in the stomach 63. When possible, the tube should preferably be inserted beyond the stomach.

EMPIRIC TREATMENT

Appropriate diagnostic microbiological procedures should be performed prior to the initiation of treatment. Although a number of recent advances have been made in the diagnosis, prevention and treatment of nosocomial pneumonia, controversies still remain and a standardized management approach does not yet exist, mainly because of diagnostic problems. The clinical and radiological signs of pneumonia can be associated with a number of causes, particularly in ventilated patients in the ICU, and the way in which these are interpreted differs between physicians. For example, if a patient presents with lung infiltrate on X-ray, fever and a raised white blood cell count, in most cases he will receive an empiric course of antibiotic treatment. If, however, there is no improvement after 48 hours and the diagnostic tests (e.g. PSB or BAL) do not suggest a bacterial etiology, some physicians would suspend antibiotic therapy and observe the patient. Other physicians may prefer to continue antibiotic therapy and increase the dosage or modify the regimen, suspecting the presence of a more resistant or unusual pathogen which was not detected. In order to guide physicians, various recommendations have been issued. These are based on a good understanding of the etiopathophysiology and of risk factors. It is important to stress that therapeutic approaches vary according to the type of patient. Only nosocomial pneumonia in ventilated patients is considered in the present article.

Early- and late-onset pneumonia in ventilated patients have different bacterial etiology. For example, early-onset pneumonia is likely to be caused by pathogens which are normal inhabitants of the oropharyngeal cavity, such as S. pneumoniac, which causes 5-10% of infections, Haemophilus influenzae, responsible for 5% of infections, or mixed flora, including anerobes. S. aureus is also a frequent cause of early-onset pneumonia. Conversely, aerobic Gram-negative bacilli, which are responsible for at least 60% of cases of pneumonia, are more likely to be involved in late-onset pneumonia. S. aureus also accounts for 20-25% of the episodes in late-onset pneumonia.

Pathogens which are more likely to arise from exogenous sources, such as P. aeruginosa, Acinetobacter spp. and methicillin-resistant S. aureus, are also usually involved in late-onset pneumonic episodes. The differences in likely causative organisms, depending on the time of onset of the pneumonic episode, should be borne in mind when choosing empiric therapy for these infections. It is also very important to take into account the epidemiology and antibiotic susceptibility pattern of each institution.

Patients with early-onset VAP and patients with no risk factors can be treated initially with agents such as β-lactams plus β-lactamase inhibitors or second- or third-generation cephalosporins (Table 6). The choice of a third-generation cephalosporin in this situation may be justified by cost consideration in certain countries, such as the USA. For patients with late-onset pneumonia, or early-onset pneumonia plus associated risk factors, such as prior antibiotics, recent hospitalization or a severe underlying condition, a broader-spectrum antibiotic is recommended, such as broad-spectrum penicillin or third-generation cephalosporin, with anti-pseudomonal activity (including broad-spectrum β-lactam/β-lactamase inhibitors), or a carbapenem. A fourth-generation cephalosporin, such as cefpirome or cefepime as single agent may be considered. An alternative therapy for patients who cannot be treated with β-lactam antibiotics is the use of an intravenous fluoroquinolone. Vancomycin may be added for initial therapy in hospitals where there is a high prevalence of methicillin-resistant staphylococci and anti-anerobic compounds added when anerobes are felt to play a predominant role.

In a recent consensus of the American Thoracic Society, two treatment schemes were recommended in this situation 1. These are based on the severity of the infection, the timing of occurrence of pneumonia (early- versus late-onset) and the presence of risk factors. The main difference from our recommendations is the use of an aminoglycoside or fluoroquinolone in combination with a β-lactam. One study has shown no difference in the success rate between treatment with imipenem alone and imipenem plus netilmicin in patients with nosocomial pneumonia and sepsis 91. It has been suggested, however, that the presence of an aminoglycoside in severe cases may improve the activity of β-lactam antibiotics which are less potent than imipenem. For certain pathogens, such as Enterobacter and Pseudomonas spp. and in severe situations, combination therapy may be superior to monotherapy 81.

Whichever treatment scheme is used, a β-lactam agent is generally included unless the patient is allergic to penicillin. This may be a β-lactam/β-lactamase inhibitor combination, a second- or third-generation cephalosporin or a carbapenem. Considering the etiology of nosocomial pneumonia and the in vitro activity of various β-lactam antibiotics, fourth-generation cephalosporins, such as cefpirome or cefepime, are worthy of consideration as alternatives to the β-lactam antibiotics currently used. These compounds have excellent Gram-positive activity, covering penicillin-resistant pneumococci and methicillin-susceptible S. aureus, as well as an increased stability and lower affinity for inducible Gram-negative β-lactamases as compared to third-generation cephalosporins 92. Although there is not yet a large amount of clinical data available on the efficacy of fourth-generation ccpnaiosporins in nosocomiai pneumonia, the Cefpirome Study Group reported on 91 patients with community-acquired or nosocomial pneumonia who received cefpirome intravenously 93. Successful clinical and bacteriological results were obtained in > 90% of patients. The authors concluded that cefpirome was an effective empiric treatment for moderate-to-severe pneumonia in hospitalized patients. However, difficult pathogens, such as Enterobacter cloacae and P. aeruginosa, were isolated in only a minority of patients. Although P. aemginosa is probably covered adequately by cefpirome while awaiting microbiological results, other agents with improved activity against this organism should be selected when P. aeruginosa is isolated.

A larger and more recent multicenter study compared the efficacy and safety of cefpirome and ceftazidime, either as monotherapy or in combination, with an aminoglycoside or metronidazole in the empiric treatment of 400 patients with nosocomial and community-acquired pneumonia in the ICU 94, Clinical failure rates at the end of treatment were 34% vs 36% for cefpirome bid and ceftazidime tid, respectively, with no difference in outcome between monotherapy and combination therapy. In a non-comparative study, a satisfactory clinical response was obtained in 89% of patients receiving cefepime bid 95. Based on these studies, it can be assumed that fourth-generation cephalosporins are good alternatives to other broad-spectrum antibiotics for the treatment of nosocomial pneumonia.

CONCLUSIONS

Nosocomial pneumonia in ventilated patients is caused by a combination of factors resulting in a variety of etiologies. Precise microbiological diagnosis requires the appropriate critical use of invasive methods and close collaboration between clinicians and microbiologists. Understanding the pathogenesis and epidemiology of infection allows the clinician not only to define optimal preventive strategies, but also to decide upon optimal empiric therapy while awaiting microbiological results.

DISCUSSION

Prof. P. Francioli: The exact burden of VA pneumonia in terms of morbidity and mortality is not well established and have been addressed specifically in only a few studies.

Prof. A. Torres: Several papers have indicated that the attributable mortality of VA pneumonia is approximately 30%. However, this is controversial. For example, Papazian et al. recently suggested that the initial microbiological diagnosis might be a confounding factor. In this study, attributable mortality was not due to VA pneumonia in itself but mostly related to the appropriateness of the initial empiric therapy.

Prof. P. Francioli: Differences in the results of various studies might also be related to the method used for the diagnosis of pneumonia. Hence, the diagnosis is the cornerstone of any study addressing the problem of VA pneumonia, but is hampered by the lack of a method which would be both highly sensitive and specific, and readily available. At the present time, lung specimens taken for microbiological and histological examination are considered to be the ‘gold standard’ technique.

Prof. A. Torres: There are problems associated with the ‘gold standard’ technique. In particular, these studies are almost exclusively post-mortem studies, and it is unlikely that a technique based on obtaining lung specimens is going to be widely available in clinical situation.

Prof. W. Wilson: Given the problems of the ‘gold standard’ for the diagnosis of pneumonia, are there any recommendations regarding which technique is the most appropriate?

Prof. J. Chastre: Several different techniques are available, but all of them are associated with methodological problems, and it remains unclear which is the most appropriate method. At this stage, BAL and PSB are considered to be the most accurate methods.

Prof. P. Francioli: Recent studies have suggested quantitative culture of endotracheal aspirates might be as accurate as more invasive methods.

Prof. J. Chastre: Yes, and this needs further investigation. It is important to emphasize that appropriate sampling should be obtained prior to initiation of antimicrobial treatment, when possible.

Prof. P. Francioli: In addition, appropriate samples should be obtained before any change in the antibiotic treatment.

Prof. P. Shah: It should be mentioned that new methods for diagnosing pneumonia have been and are being investigated, such as measurement of the endotoxin or procalcitonine levels in bronchial secretions. These methods have shown promising results.

Prof. P. Francioli: In recent years, progress in the prevention of VA pneumonia has been made. Environmental sources of infection have certainly been reduced by the systematic use of adequate hygienic measures. New devices such as endotracheal tubes allowing the aspiration of subglottic secretions have been developed.

Prof. A. Torres: Undoubtedly, these tubes appear to bring some progress. However, they have their own problems, such as contamination or blockage of the lumen.

Moreover, studies were performed with prototypes and should probably be confirmed with the tubes which are now becoming available on the market.

Prof. P. Francioli: Two preventive measures are still controversial: the choice of anti-stress ulcer prophylaxis regimen and the selective digestive decontamination (SDD). Regarding the anti-stress ulcer prophylaxis regimen, meta-analysis of the existing studies suggest that regimen which elevate the gastric pH might be associated with increased incidence of VA pneumonia. However, the most controversial measure is probably SDD.

Dr. M. Langer: It is true that this measure is the most controversial, but this is more of an emotional than rational reaction. Hence, the data are very clear: there is no doubt that certain SDD regimens decrease the incidence of VA pneumonia. This is also confirmed by several meta-analysis. However, it is also true that no firm conclusion can be drawn regarding the effect of SDD on mortality. More studies are needed to identify which patients might benefit most from such a measure. The issue is emotional and controversial because of the risk of selecting resistant microorganisms. This risk might depend on the type of regimen used. More work is needed in this area.

Prof. P. Shah: At the present time, one indication for the use of SDD might be an epidemic situation related to a multi-resistant microorganisms.

Prof. P. Francioli: Regarding the initial empirical treatment of VA pneumonia, it is important to distinguish early-onset pneumonia from late-onset pneumonia.

Prof. J. Acar: Could you clarify this concept?

Prof. M. Langer: We were the first to describe the difference in the spectrum of microorganisms responsible for a pneumonia in relation to the time of onset of infection. In an Italian multicenter study, approximately 50% of the pneumonic episodes developed within the first 4 days after intubation. Most were due to bacteria recovered from the normal oropharyngal flora. In contrast, patients developing pneumonia at a later stage were most likely to be infected with Gram negative bacilli. The same was true for patients with early onset pneumonia and risk factors, such as patients with prior antibiotic treatment. This was confirmed by several other studies and fits with our present knowledge of the pathogenesis of VA pneumonia. This distinction is important for the treatment. Thus, initial empirical treatment of pneumonia will depend on the time of onset of the episode and the presence or not of risk factors.

Prof. J. Acar: Anerobes are also responsible for episodes of early-onset pneumonia.

Prof. M. Langer: This is particularly true when gross aspiration has occurred. In addition, some studies have showed that anerobes can be recovered in approximately 30% of pneumonic episodes. The exact role of this anerobes remain to be determined.

Prof. P. Francioli: In patients with early-onset pneumonia and no risk factors, the first line regimen would be a β-lactam/β-lactamase inhibitor combination or a second- generation cephalosporin. Should a third-generation cephalosporin (without antipseudomonal activity) also be recommended for this indication?

Prof. W. Wilson: Second-generation cephalosporin are unavailable in certain countries and are also more expensive than third-generation cephalosporins in many countries. This is true for the USA. Thus, third-generation cephalosporins should be included as alternative to second-generation cephalosporins in the treatment of early-onset pneumonia with no risk factors. Regarding the β-lactam/β-lactamase inhibitor combinations, it should be mentioned that amoxicillin-clavulanate is more active than ampicillin-sulbactam against penicillin resistant pneumococci. However, parenteral amoxicillin-clavulanate is not available in the USA, where only ampicillin-sulbactam is commercialized.

Prof. J. Acar: There are some data suggesting that nosocomial Gram-negative pneumonia may also occur after the first few days of intubation.

Prof. A. Torres: This is true, but these patients generally have risk factors, mostly prior antibiotic treatment or a severe underlying disease.

Prof. P. Francioli: Initial empirical therapy in these patients should be identical to that of patients with late-onset pneumonia.

Prof. J. Chastre: We should emphasize that, whatever patient is considered, appropriate microbiological sampling should be performed before the initiation of empirical treatment. Moreover, empirical treatment should be adapted according to the microbiological results and the clinical response.

Prof. P. Francioli: For late-onset pneumonia or for patients with early onset-pneumonia and risk factors, a broader coverage is needed as compared to early-onset pneumonia. This is due to the fact that the distribution of microorganisms is different and involves certain Gram-negative microorganisms such as the Pseudomonas, Enterobacter or Acinetobacter which are usually resistant to the antibiotics recommended for early-onset pneumonia. A broad-spectrum β-lactam with antipseudomonal activity or a carbapenem are the drugs of choice. In penicillin allergic patients, a fluoroquinolone combined with or without an anti-anerobe compound is an alternative.

Prof. J. Acar: Third-generation cephalosporins with anti-pseudomonal activity carry the risk of encountering Enterobacteriacae producing inducible cephalosporinase and selecting for such resistant microorganisms. One should probably select agents such as cefpirome or cefepime, which may avoid the selection of resistant microorganisms.

Prof. W. Wilson: The American Thoracic Society recommends combination therapy for empirical treatment of VA pneumonia, mainly the combination of a broad-spectrum β-lactam (e.g. a third-generation cephalosporin) and an aminoglycoside.

Prof. P. Francioli: This is still a controversial issue. Some studies have demonstrated that adding gentamicin or netilmicin to imipenem has not improved the outcome of patients with nosocomial pneumonia. Whether or not this is true for agents other than carbapenems remains to be determined. However, some subgroups of patients might benefit from combination therapy, in particular those with Pseudomonas, Enterobacter or Acinetobacter infections. For the other cases, a carbapenem, a fourth-generation cephalosporin or piperacillin-tazobactam are probably adequate first-line empiric therapy for late-onset pneumonia.

Prof. P. Shah: A combination therapy should probably be envisioned in the initial regimen of all severe cases.

Prof. P. Francioli: Glycopeptides should be added to the treatment regimen according to the local epidemiology of MRSA infections. This is particularly true if Gram-strain of the bronchial specimen shows Gram-positive cocci.

Dr. M. Langer: It should be remembered that S. aureus is also a frequent pathogen encountered in early-onset pneumonia. However, in this instance, S. aureus has been community-acquired and is generally susceptible to methicillin.

Prof. P. Francioli: This discussion illustrates that there are still many open questions in the field of ventilator-associated pneumonia.