Introduction

Ventilator-associated pneumonia (VAP) is the most common nosocomial infection among intensive care unit (ICU) patients [1]. Debate persists about the mortality attributable to VAP, but VAP causes substantial morbidity by increasing the ICU stay [2, 3, 4, 5, 6, 7]. Risk factors for VAP have been described, but few investigations have been led in specific populations [8, 9, 10, 11].

Trauma admission is a recurrent risk factor for VAP [9, 12, 13, 14, 15]. The incidence of VAP among trauma patients is high, because of pharyngeal aspiration in patients with depressed consciousness. Risk factors for nosocomial pneumonia include injury severity score (ISS) >20, head injury, emergent intubation, collapse, and blunt trauma [14, 15, 16, 17]. Moreover, trauma patients undergo transport for various procedures [18]. The supine position for prolonged periods, the manipulation of their ventilator circuits, and the aspiration of contaminated secretions accumulated above the cuff of the tracheal tube also represent risk factors for VAP [18].

Selective digestive decontamination (SDD) has been extensively studied for preventing VAP [19, 20, 21, 22]. The SDD regimen consists of topical non-absorbed antibiotics applied orally and through a nasogastric tube, and is associated with a short course of parenteral antibiotic [19]. SDD in trauma patients leads to decreased numbers of VAP episodes [15, 21]. This research was presented during the 12th Congress of the European Society of Anaesthesiology [23].

The risk factors for late-onset VAP in severe trauma patients receiving SDD have never been investigated. In the present study, we evaluated the independent factors associated with the occurrence of late-onset VAP in severe trauma patients receiving SDD.

Patients and methods

This prospective study was conducted from January 1998 to 2002 in a 700-bed teaching hospital. The protocol was in accordance with the ethical standards of our hospital’s Committee for the Protection of Human Subjects. Informed consent was not obtained because this study did not modify existing diagnosis or therapeutic strategies. The ICU has 16 beds admitting medical, surgical, and trauma patients. To be eligible for evaluation patients had to have an ISS >15, and duration of mechanical ventilation >5 days. A team consisting of a senior physician with a nurse managed patients on the field. Criteria for intubation were Glasgow Coma Scale score (GCS) <9, acute respiratory distress, or agitation. The use of neuromuscular blockers for intubation, as well as the choice of agents, was dependent on the decision of this team.

All ventilated patients were monitored daily for the development of VAP. Age, sex, simplified acute physiology score (SAPS) II, ISS, GCS after resuscitation, and sepsis-related organ failure assessment (SOFA) scale were recorded on admission. The following dates were recorded: admission and discharge from the ICU and the hospital, onset of pneumonia. ICU and hospital lengths of stay, mortality rate at the time of discharge from ICU and hospital, and duration of mechanical ventilation were calculated. All events and interventions occurring until the ICU discharge were collected.

All patients received SDD as described elsewhere [22]. SDD agents consisted on polymixin E, gentamicin, and amphotericin B. A 2% mixture of these drugs in Orabase was applied four times daily on the oral mucosa and in the digestive tract until extubation. For the first 3 days, systemic cefazolin (1 g × 3 daily) was given to SDD patients.

Stress-ulcer prophylaxis was used in patients with a history of ulcer. Sucralfate was recommended in our protocol, but histamine-2 receptor antagonist was kept on in patients already treated with this drug (two patients). Patients received early enteral nutrition except those with a bowel perforation. They were in semi-recumbent position, and underwent mechanical ventilation with the ventilator set in volume-controlled mode. Heat and moisture exchanger bacterial filters (Gibeck, Stockholm, Sweden) were inserted between the endotracheal tube and the ventilator circuit except for patients with acute respiratory distress syndrome (ARDS). Boluses of cistracurium were administered to assist in the treatment of patients with increased intracranial pressure (ICP), to facilitate mechanical ventilation, and to decrease oxygen demand in patients with ARDS [24].

Tracheal colonization was defined by the presence of potentially pathogenic microorganisms, including Staphylococcus aureus, Streptococcus pneumoniae, Haemophilus influenzae, Enterobacteriaceae, Pseudomonas aeruginosa, and Acinetobacter spp. in the tracheal aspirate obtained upon admission in the ICU and then twice weekly [13].

The diagnosis of VAP was established when the following criteria were fulfilled: 1) bronchial purulent sputum; 2) body temperature >38 °C or <36 °C; 3) worsening of arterial oxygenation; 4) white blood cells >12,000/mm3 or <4,000/mm3; 5) chest radiograph showing new or progressive infiltrates; and 6) presence of at least one microorganism at a concentration of at least 104 colony-forming units/ml on the broncho-alveolar lavage [25].

The antimicrobial therapy was administered according to our local guidelines. Empirical treatment was administered after the collection of microbiological samples. For early-onset nosocomial infections, ceftriaxone was selected. Piperacillin-tazobactam was used for patients with peritonitis. For late-onset nosocomial infections, cefepime or imipenem was selected. An aminoglycoside or a quinolone was added according to the clinical impression of the patient’s treating senior physician. Vancomycin was prescribed when methicillin-resistant staphylococci were suspected. De-escalation of antibiotic therapy consisted of either deleting one of the antibiotics of the prescribed combination or, whenever possible, to use a beta-lactam with a narrower spectrum.

Prior antimicrobial treatment was defined as the treatment of a documented or suspected infection before the occurrence of the late-onset VAP. Antibiotics administered for SDD, including cefazolin, and those for surgical prophylaxis, were excluded from this definition.

The following criteria were assessed as potential risk factors: 1) qualitative criteria related to past medical history: age, gender, diabetes, smoking, chronic alcoholic intoxication, immunodepression, pre-existing organ dysfunction; 2) qualitative criteria available on admission: intubation on the scene or in ICU, oral or nasal intubation, surgery on admission, vasopressors, type of trauma, head trauma, type of head trauma, chest trauma, type of chest injury, abdominal trauma, association of chest and abdominal trauma, tracheal colonization, neuromuscular blockers for intubation, mean arterial pressure <70 mmHg for longer than 5 min; 3) quantitative criteria on admission: SAPS II, ISS, SOFA, number of vasopressors, number of rib fractures, alcohol blood concentration; 4) qualitative criteria available during the ICU stay: ICP monitoring, mydriasis, ICP >25 mmHg, osmotherapy, cerebrospinal fluid derivation, use of propofol, hyperventilation, craniectomy, positive end-expiratory pressure (>5 cmH2O), tracheostomy, vasopressor, transfusion (>4 red blood cell packs), nasogastric tube, sedation (midazolam, ketamine, propofol, pentobarbital, sufentanil), neuromuscular blockers, steroids, stress-ulcer prophylaxis, chest tube, organ dysfunction, prior use of antimicrobial treatment except SDD, ARDS; and 5) quantitative criteria during the ICU stay: number of intubations, number of catecholamines, duration of sedation, duration of chest tube, number of organ dysfunction, duration of intubation, mechanical ventilation, ICU stay, central venous line placement, arterial catheter placement, antimicrobial treatment except SDD.

Statistical analysis was performed using the Statistical Analysis software package (version 5, SAS Institute, Cary, N.C., USA). Univariate analysis was conducted to determine potential risk factor of VAP occurrence. The χ2 or Fisher’s exact test was used for qualitative variables, and Student t-test was used for quantitative variables. The required significance level was set at a P-value less than 0.05. Multivariate analysis quantified the respective effect of each variable on the occurrence of late-onset VAP. Stepwise logistic regression was performed (forward method, likelihood ratio). Explanatory variables in the logistic regression were: a) variables identified as potential risk factor by the univariate analysis with a cut off at 0.2; and b) variables known as risk factors by the scientific community. The condensed model was presented with crude odds ratio and 95% confident interval.

Results

Of the 159 severe trauma patients requiring at least 5 days of mechanical ventilation, 90 (56%) exhibited 116 late-onset VAP episodes on day 10.7±6.1 days after admission [range, 5–40 days] (Table 1). The main pathogens involved in late-onset VAP were H. influenzae (32%), S. aureus (31%), and S. pneumonia (13%), Escherichia coli (12%), and A. baumanii (7%). Among Gram-positive bacteria, 6% of S. aureus were methicillin-resistant, whereas 7% of Gram-negative bacteria were multi-resistant to antibiotics. P. aeruginosa was isolated in two patients. Prior antimicrobial treatment was administered in 58 patients with early-onset ventilator-associated pneumonia (20), intra-abdominal sepsis (18), catheter-associated urinary tract infection (6), and miscellaneous infections (14). Prior tracheal colonization was observed in 119 (75%) patients, and 77 (65%) among them developed a late-onset VAP. The same bacteria were found in colonization and infection in 53 patients.

Table 1. Characteristics of patients on admission with P-value comparing patients with and without late-onset ventilator associated pneumonia. (SAPS Simplified Acute Severity Score, ISS Injury Severity Score, GCS Glasgow Coma Scale, SOFA Sepsis-related Organ Failure Assessment scale).
Full size table

The significant results of univariate analysis related to the qualitative events are reported in Table 2. Age, sex, significant past medical history of chronic alcohol abuse or smoking, immunosuppression, the place where intubation has been performed, the route of intubation, the need for surgery at admission, the type of injury, the severity of illness according to SAPS II, ISS, SOFA scores, the need for vasopressors, and alcohol intoxication were not found to be associated with an increased risk for late-onset VAP (Table 3). The criteria related to the ICU stay are collected in Fig. 1. The univariate analysis showed that the duration of intubation, mechanical ventilation, ICU stay, need for central line and arterial catheter, and prior antimicrobial treatment (excluding SDD) were significantly increased in patients with late-onset VAP compared with those without late-onset VAP. On the contrary, neurological events (ICP monitoring, osmotherapy, hyperventilation) were not observed more often in these patients. The following events were not related with a significant increase in late-onset VAP: catecholamine, transfusion of red blood cells, tracheotomy (n = 5), positive expiratory pressure >5 cmH2O (n = 41), chest tube (n = 42). Interventions like sedation (n = 134) including neuromuscular blockers (n = 48), steroids (n = 14), stress ulcer prophylaxis (n = 38), re-intubation (n = 62) were not associated with an increased risk for late-onset VAP.

Table 2. Univariate analysis: significant qualitative criteria associated with the occurrence of late-onset ventilator-associated pneumonia.
Full size table
Table 3. Univariate analysis. Factors not related to the development of late-onset ventilator-associated pneumonia. (SAPS Simplified Acute Severity Score, ISS Injury Severity Score, GCS Glasgow Coma Scale, SOFA Sepsis-related Organ Failure Assessment scale).
Full size table
Fig. 1.
figure 1

Duration of intubation, mechanical ventilation, ICU stay, central line catheterization, arterial catheterization, and antibiotherapy treatment in patients with and without late-onset ventilator-associated pneumonia. *P<0.05

Full size image

The condensed model is presented in Table 4. The multivariate analysis identified the following factors: prior tracheal colonization, duration of intubation, length of ICU stay, and the use of non-depolarizing neuromuscular blockers for intubation. Exposure to prior antimicrobial treatment [odds ratio, 0.31 (CI, 0.12–0.74)] conferred a protection.

Table 4. Multivariate analysis: criteria associated with the occurrence of late-onset ventilator-associated pneumonia.
Full size table

The ICU mortality was of 30%. There is no significant difference between patients with late-onset VAP (26%) and those without late-onset VAP (37%).

Discussion

We examined the risk factors for late-onset ventilator-associated pneumonia, occurring in 56% of a cohort of 159 severe trauma patients receiving SDD. It is essential to emphasize that 88% of our patients presented a GCS score below 9 and 53% a chest trauma; both are two well-described risk factors for early-onset VAP [12, 13, 14, 17]. Patients with head trauma have compromised local airway immune defense mechanisms [12]. Traumatic injuries create a state of relative immunosuppression [26]. Moreover, the chest X-ray interpretation is complex in this context, which could have increased the number of positive diagnosis. Thus, the high incidence can be explained by the selection of patients with head and chest injuries requiring prolonged mechanical ventilation [17]. However, one hypothesis may be that the SDD delayed the outbreak of VAP, resulting in an increase number of late-onset VAP.

The use of neuromuscular blocking agents is not associated with an increased risk for late-onset VAP in our study. Nevertheless, non-depolarizing agents are linked with a 3.4-fold increase in late-onset VAP occurrence. Depolarizing agents tend to be associated with a protective effect. There is no strong recommendation concerning the use of neuromuscular blockers for the endotracheal intubation of trauma patients [27]. One hypothesis is that short-acting agents provide good conditions for intubation, which results in a decrease of the amount of aspiration [28]. However, suxamethonium induces a transient rise of the ICP [29]. Regarding neuromuscular blockers for intubation, our results suggest the need for future investigations.

The present study highlights the protective effect of prior antimicrobial treatments on the occurrence of late-onset VAP, as previously described by Cook et al. [9]. Actually, antibiotic administration was related to both an increased and a decreased risk for VAP [30]. The bacteria causing the late-onset VAP are typical of early-onset pathogens in 74% of our patients, with a low rate of multiresistant bacteria. This confirms the moderate effect of SDD on microbial ecology [22], and could be related to our protocols recommending short duration of treatment for ventilator-associated pneumonia. This statement should be confirmed by a rigorous study. However, as highlighted by others, the use of prophylactic antibiotic for selective digestive decontamination purpose is not associated with an increase of bacterial resistance [31, 32].

The relationship between the tracheal colonization and VAP has been demonstrated [12]. Both the upper airways and stomach represent independent reservoirs for tracheal colonization with ICU-acquired pathogens and VAP. Tracheal colonization within 24 h of intubation is an independent risk factor for VAP in patients with head injury [13]. The selective decontamination of the subglottic area in mechanically ventilated trauma patients significantly reduced tracheal colonization [33]. However, the use of SDD is never followed by a complete eradication of tracheal colonization. Colonization is significantly reduced, but still persistent, although at much lower levels. In the study of Bergmans et al., the use of SDD decreased colonization by 50% (52% in the control group versus 22% in the treated group) [34]. Bonten et al. observed a persistent or acquired colonization in 28–50% of patients receiving SDD. The persistence of tracheal colonization could be of concern, since it may lead to the subsequent development of nosocomial VAP [35]. However, in the present study, early tracheal colonization increased the risk for late-onset VAP by only 0.03. The hypothesis could be a persistent colonization with a low inoculum of bacteria which interferes itself with the emergence of late-onset pathogens. The SDD could induce a delay in the development of VAP with these persistent bacteria, explaining the number of early-onset pathogens causing late-onset VAP.

In agreement with other studies, ICU length of stay and duration of intubation are independent risk factors for VAP [5, 9]. Long-term intubation is related to prolonged mechanical ventilation due to head trauma. Regression analysis identified the length of intubation as an independent risk factor instead of head trauma or prolonged mechanical ventilation. We did not collect data on body position, but a causal relationship between supine position and VAP have been documented in many studies [30, 36, 37, 38]. For instance, transport in the supine position to the site of the radiological procedure could be linked to an increased risk for VAP, but this issue was not examined [39].

Some known risk factors were not found in our patients receiving SDD. The need for re-intubation has been associated with VAP [8]. In the present study, a sample-size effect may limit the effect of this risk factor. Chest trauma has been related to a 3.1 increase of the risk for early-onset VAP [16]. A combined abdominal and thoracic trauma represents the strongest risk factor for early-onset pneumonia [17]. This is not confirmed in our study since neither chest nor abdominal trauma are associated with late-onset VAP. Similarly, factors reflecting the severity of trauma, i.e. hypotension, ARDS, and GCS score are not found as independent risk factors. Moreover, SAPS II, ISS, and SOFA are not reliable to predict late-onset pneumonia. Probably, the most severely ill patients develop an early-onset VAP, with a protective effect of antibiotics on the occurrence of late-onset VAP. However, in a study focusing on early-onset pneumonia, the APACHE II score was not significantly different between patients with and without VAP [17]. The occurrence of a late-onset VAP is not associated with an increase of mortality, which confirms the results of previous studies [2, 40]. A significant reduction of mortality has been demonstrated in mixed medical and surgical patients receiving SDD [31].

In conclusion, this study investigates the risk factors for late-onset VAP in trauma patients receiving SDD. As we previously demonstrated, the occurrence of a VAP induces an increase in the duration of mechanical ventilation and ICU stay. In trauma patients receiving SDD, several independent risk factors for the development of late-onset VAP are similar to those found in other patients: duration of intubation, length of ICU stay, and prior tracheal colonization. Some other independent risk factors were not found: re-intubation, chest trauma, abdominal trauma, hypotension, ARDS, and low GCS. The use of neuromuscular blocking agents is not associated with the risk for late-onset VAP and the type of neuromuscular blocking agents should be evaluated in future studies by a rigorous protocol.