INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients with traumatic and medical head injury have been shown to be at particularly high risk of ventilator-associated pneumonia (VAP). Its incidence is estimated to reach 40 to 50% (1). The most frequent etiologic agents found include Staphylococcus aureus and, less frequently, Steptococcus pneumoniae and Hemophilus influenzae (4, 5, 8, 9). The early onset of pulmonary infection and the peculiar microbial pattern may be due to oropharyngeal or gastric colonization followed by high inoculum aspiration of oropharyngeal secretions. Patients may aspirate oropharyngeal secretions shortly after brain injury, during resuscitation, and as a consequence of intubation. Although the predominant pathogens of early onset pneumonia in patients with head injury have been well established in several epidemiologic studies, the precise relation of prior upper airway, tracheobronchial, and gastric colonization patterns with the development of pneumonia and microbial patterns has not been settled. Moreover, the relative homogeneity of a population with a high incidence of early onset pneumonia that in turn is equally at risk for late onset pneumonia gives the opportunity to study the differences and similarities in the pathogenesis of early- and late-onset pneumonia.

Thus, the aims of the study were to describe the qualitative and quantitative colonization patterns in patients with surgical and medical head injury, to search for risk factors of initial and follow-up bacterial colonization at different sites of the upper and lower airways as well as the stomach, and finally to determine the incidence and microbial etiology of VAP and its relation to prior colonization patterns.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Population

We prospectively studied 48 patients admitted at two different medical and surgical ICUs at a 1,000-bed teaching hospital during a 16-mo period. Patients were enrolled because of head trauma or medical stroke. All patients were orally intubated from the beginning of the study. No patient had tracheostomy. The time lapse between intubation and the beginning of the study protocol was  24 h in all cases. Selective digestive decontamination (SDD) was not administered in any case. Subglottic drainage was not applied to any patient. Exclusion criteria were severe immunosuppression (organ transplantation, neutropenia < 1 × 10⁹/L, AIDS) and evidence of pulmonary infection or suspicion of gross aspiration at admission. The study was approved by the local Ethical Committee. Informed consent was obtained from the nearest relatives of every patient.

Data Collection

Demographic, clinical, and treatment data were collected from all patients. These included: age, sex, type of cerebral injury, comorbid illnesses, McCabe's prognostic score, antimicrobial therapy, simplified acute physiologic score (SAPS), Glasgow Coma Score, Pa_O2/FI_O2 at admission, the initial presence of shock, prophylaxis and treatment of cerebral edema, presence of intracranial pressure monitoring, type of stress ulcer prophylaxis, type of nutrition, days of mechanical ventilation, and outcome.

Study Protocol

Within the first 24 h after admission to the ICU and intubation, nasal and pharyngeal swabs (NS and PS) (Medio Amies, Barcelona, Spain), tracheobronchial aspirate (TBAS) (Mocstrap; Productes Clinics S.A., La Llagosta, Barcelona, Spain) as well as 3 ml of gastric juice (GJ) were obtained. Next, flexible bronchoscopy was performed with protected specimen brush (PSB) sampling from either lower lobe. NS, PS, TBAS, and GJ samples were taken daily until the fourth day of ICU admission, and every 72 h thereafter, until the end points extubation, development of suspected VAP, or death. In case of death, necropsy was done when possible.

Chest radiographs were taken daily. Patients with suspected pneumonia were investigated by TBAS and flexible bronchoscopy with PSB in the area most prominently affected. In addition, blood cultures were obtained.

The preparation of all samples was performed as described previously (10). All samples were quantitatively plated on blood, chocolate, Wilkins-Chalgren, and Sabouraud agar media (serial dilutions of 1:10, 1:100, and 1:1,000). If negative, the plates were discarded after 3 d of testing for aerobic bacteria and after 4 wk of testing for fungi. If positive, counts of colony-forming units (cfu) per milliliter and identification as well as susceptibility testing using standard methods were done for the microorganisms (11). Bacterial (and fungal) indices were calculated using a formula previously described (12).

Definitions

Microorganisms. Microorganisms were classified into potentially pathogenic microorganisms (PPMs) and non-PPMs. PPMs were those recognized as causing respiratory infections: Gram-positive cocci such as Streptococcus pneumoniae and Staphylococcus aureus, Gram-negative cocci such as Moraxella catarrhalis, and Gram-negative rods such as Hemophilus influenzae, Enterobacteriaceae and Pseudomonas aeruginosa. Non-PPMs were those microorganisms usually not involved in respiratory infections in the nonimmunosuppressed host (Streptococcus viridans group, coagulase-negative staphylococci, Neisseria spp., Corynebacterium spp., and Candida spp.) (13).

Colonization. The presence of bacterial microorganisms in the upper airways (as determined in NS or PS) or the stomach (as determined in GJ) in at least one sample in any amount was considered as colonization. In addition, the presence of bacterial microorganisms in the tracheobronchial tree (as determined by TBAS) and lower airways (as determined in PSB) in at least one sample in any amount was considered as colonization in the absence of confirmed pneumonia (11). PPMs and non-PPMs were analyzed separately. For purpose of analysis, Gram-positive isolates (Streptococcus pneumoniae and methicillin-sensitive Staphylococcus aureus [MSSA]) and Hemophilus influenzae were grouped together (Group I pathogens) and opposed to the group of Gram-negative aerobic bacilli (GNEB), Pseudomonas aeruginosa, and Acinetobacter spp. (Group II pathogens).

Pneumonia. Pneumonia was clinically suspected upon the presence of new and/or progressive pulmonary infiltrates on chest radiograph in addition to two of the following criteria: fever  38.3° C or hypothermia < 36° C, leukocytosis  12 × 10⁹/L, and purulent tracheobronchial secretions. Pneumonia was considered as definite in the presence of (1) a colony count  10³ cfu/ml of one or more PPMs from cultures of PSB or  10⁵ cfu/ml from cultures of TBAS; (2) a PPM in blood culture in the absence of an extrapulmonary focus; (3) a response to antimicrobial therapy in the absence of an alternative diagnosis; or (4) histologic evidence of pneumonia (consolidated foci and polymorphonuclear leukocyte accumulation in bronchioles and adjacent alveoli) in necropsy. PPMs (but not non-PPMs) recovered above the defined cutoffs for the confirmation of pneumonia were considered as causative microorganisms. Non-PPMs were analyzed separately. Early-onset pneumonia was diagnosed when pneumonia developed within the first 4 d of mechanical ventilation, and late-onset pneumonia when developing after the fourth day of mechanical ventilation.

Brain lesions. Brain lesions were diagnosed when head trauma (open or closed, needing surgery or not) or medical stroke were present. A Glasglow coma score < 9 was defined as coma (14).

Antibiotic treatment. Previous (short-term) antibiotic treatment was defined as the in-hospital administration of any dose of any antibiotics prior to the first sampling of respiratory and gastric secretions. Prolonged antibiotic treatment was defined as the in-hospital administration of antibiotics for more than 24 h.

Treatment with barbiturates. Treatment with barbiturates was defined as inravenous perfusion of sodium thiopental in a dose  1 mg/ kg/h for controlling intracranial hypertension.

Statistical Analysis

Results are expressed as mean ± SD if not indicated otherwise. Student's t test for independent and paired continuous variables and the chi-square test or Fisher's exact test were used when appropriate to compare proportions.

In order to detect risk factors for colonization, the following factors were tested for an association with initial and follow-up colonization rates with Group I and Group II pathogens: SAPS (< 15 or  15, according to the mean value of the whole population), Glasgow Coma Score (< 9 or  9, according to the definition of coma), as well as treatment with barbiturates (yes/no), previous (short-term) antibiotic treatment (yes/no) (for initial and follow-up colonization), prolonged antibiotic treatment (for follow-up colonization), enteral nutrition, stress ulcer prophylaxis with H₂-blockers, and intracranial pressure monitoring. Moreover, the importance of upper airways and stomach for follow-up tracheobronchial colonization was also assessed by multivariate stepwise forward logistic regression models including follow-up tracheobronchial colonization as an independent variable and initial and follow-up upper airway and gastric colonization as covariates.

Initial and follow-up colonization rates as well as enteral nutrition, treatment with H₂-blockers, previous (short-term) antibiotic treatment, prolonged antibiotic treatment, and duration of mechanical ventilation were tested for an association with early- and late-onset pneumonia. Variables significantly associated with pneumonia were included in a multivariate stepwise forward logistic regression. Time-to-event tables for early-onset pneumonia were calculated according to the Kaplan-Meier method, with log-rank test for significance of differences.

All p values are two-sided, and the level of significance was set at 5%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Forty-eight patients (28 men, 20 women) were included in the study. Mean age was 48 ± 21 yr. Brain lesions included: medical causes, 23 (48%) (spontaneous cerebral hemorrhage, n = 9 and stroke, n = 14); surgical causes, 25 (52%) (head trauma: cerebral concussion or contusion, n = 7; cerebral edema, n = 2; hemorrhage, n = 16, including 12 patients with polytrauma). Coma was present in 40 patients (20 surgical, 20 medical). The main features of the study population are listed in Table 1. The mean number of microbiologic investigations performed in each patient was 4 ± 2 (range, 1 to 6).

Antibiotic Treatment

Previous (short-term) antibiotic treatment in 22 patients (44%) consisted of intravenously administered monotherapy with second (n = 8) and third (n = 4) generation cephalosporins and clindamycin (n = 1) as well as combination regimens with third generation cephalosporins, aminoglycosides, and clindamycin (n = 2), second generation cephalosporins or penicillin, or clindamycin and aminoglycosides (n = 1 each), third generation cephalosporins and clindamycin (n = 1), first generation cephalosporin, penicillin, and aminoglycoside (n = 1), and third generation cephalosporins, penicillin, and vancomycin (n = 1), and others (2).

Prolonged antibiotic treatment was given in 18 patients (38%) for a mean of 5 ± 3 d and included monotherapy with quinolones (n = 1), third generation cephalosporins (n = 6) as well as combination regimen with third generation cephalosporins and aminoglycosides (n = 5), clindamycin (n = 1), aminoglycosides and clindamycin (n = 3), and clindamycin and aminoglycosides (n = 1), penicillin and vancomycin (n = 1).

Initial Colonization

The incidence of initial colonization with PPMs in at least one sample was 39/47 (83%). Colonization rates in the upper airways (NS or PS: 26/46, 57%) and tracheobronchial tree (TBAS: 23/41, 56%, and PSB: 21/36, 58%) were similar. The stomach (GJ) was colonized in 16/36 (44%) of patients. The total and local colonization rates were not significantly different when comparing medical with surgical patients.

Initial colonization in NS, PS, TBAS, and PSB frequently accounted for Group I pathogens (NS or PS: 20/46, 44%; TBAS: 20/41, 49%; PSB: 15/36, 42%). In contrast, GJ showed these pathogens in only 4/36 (11%) cases (NS/PS, TBAS, PSB versus GJ, p < 0.005 for all comparisons). The reverse trend was found for Group II pathogens (NS or PS: 9/46, 16%; TBAS: 4/41, 10%; PSB: 7/36, 19%) and GJ: 14/36, 39% (NS/ PS, TBAS, PSB versus GJ, p = 0.05, p < 0.005, p = 0.07, respectively) (Figure 1). There were no significant differences between medical and surgical patients. The frequencies of bacterial isolates are listed in Table 2.

View larger version (27K): [in this window] [in a new window]   Figure 1.   Comparison of initial and follow-up colonization rates with Group I and Group II pathogens at different sites. NS = nasal swab; PS = pharyngeal swab; TBAS = tracheobronchial aspirate; PSB = protected specimen brush; GJ = gastric juice. Solid bars = initial colonization rate in %; dotted bars = follow-up colonization rate in %. Group I pathogens: p = NS for all comparisons. (See METHODS for definitions of Group I and II pathogens.)

Follow-up Colonization

Follow-up cultures were performed in 43 patients. Of these, 39 (91%) had colonization with PPMs in at least one sample. The colonization rates at the upper airways and TBAS increased significantly as compared with the initial investigation (NS or PS: 37/43, 86%, p < 0.01; TBAS: 33/43, 77%, p = 0.04), with a similar trend for GJ (25/42, 60%, p = 0.18). Colonization rates were not significantly different when comparing medical with surgical patients.

The colonization rates for Group I pathogens were similar in the upper airways (NS or PS: 24/43, 56%; TBAS: 19/43, 44%) and GJ: 3/42, 7% as compared with the initial cultures. The Group I colonization rates of NS/PS and TBAS continued to be significantly different as compared with GJ (p < 0.0001 for both comparisons). However, the rates of Group II pathogens were significantly higher in the upper airways (NS or PS: 25/43, 58%, p < 0.001; TBAS: 21/43, 49%, p < 0.0001; and in GJ: 25/42, 60%, p < 0.01) (Figure 1). The colonization rates of NS/PS and TBAS were no longer different as compared with GJ (p = 0.89 and 0.32, respectively). Thus, the increase in colonization rates during follow-up was mainly due to an increase of colonization with Group II pathogens. This trend was uniformly observed in medical and surgical patients. The frequencies of bacterial isolates are listed in Table 2.

Time Course of Follow-up Colonization

In patients with negative initial and positive follow-up cultures, the mean time to colonization was 48 ± 44 h (n = 7) for Group I pathogens and 56 ± 45 h (n = 19) for Group II pathogens in the upper airways (NS or PS), 58 ± 27 (n = 5) and 43 ± 32 h (n = 15) in the tracheobronchial tree (TBAS) and 24 (n = 1) and 44 ± 35 h (n = 12) in GJ, respectively. Similarly, in patients with positive initial cultures with Group I pathogens but absence of Group II pathogens, the mean time to follow-up colonization with Group II pathogens was 40 ± 21 h in the upper airways (NS or PS) (n = 9), 36 ± 24 h in the tracheobronchial tree (TBAS) (n = 4), and 24 ± 0 h in GJ (n = 2).

Quantitative Colonization Patterns

The mean bacterial indices of both Groups I and II pathogens in nose, pharynx, and trachea showed a similar evolution. The initial Group I bacterial load (log₁₀ Bacterial Index) of the nose and trachea was significantly higher as compared with the initial Group II load (2.1 ± 0.5 versus 0.5 ± 0.2, p < 0.001 and 3.4 ± 0.7 versus 0.6 ± 0.3, p < 0.0001, respectively), with a similar trend for the pharynx (1.7 ± 0.4 versus 0.8 ± 0.3, p = 0.1). After deflection points (between Days 2 and 4), Group II bacterial loads turned out to be significantly higher in the nose (0 versus 8.1 ± 1.8), pharynx (0 versus 9.9 ± 1.6), and trachea (0.7 ± 0.7 versus 10.1 ± 5.1) on Day 6 (p < 0.01, each). Conversely, there was a significantly higher initial Group II bacterial load in gastric juice (0.6 ± 0.3 versus 2.8 ± 0.7, p < 0.01), whereas the Group I bacterial load dropped to zero by Day 4 (Figure 2).

View larger version (16K): [in this window] [in a new window]   Figure 2.   Quantitative colonization patterns in the nose, pharynx, trachea, and stomach. Log10 BI = mean sum of logarithms of colony counts, i.e., mean bacterial index (BI). Data are presented as mean ± SEM. Note that Days 5 and 6 of sampling correspond to Days 7 and 10 of mechanical ventilation. Number of samples on days of sampling, 1 to 6; nose: 43, 40, 33, 23, 17, 5; pharnyx: 44, 40, 31, 25, 17, 4; trachea: 41, 37, 28, 23, 17, 6; stomach: 36, 40, 30, 25, 14, 5. Dotted lines = Group I pathogens; solid lines = Group II pathogens. (See METHODS for definitions of Group I and II pathogens.)

Risk Factors for Bacterial Colonization

Glasgow Coma Score < 9 was associated with an initial upper airway colonization (NS or PS) with Group I pathogens. Previous (short-term) antibiotic treatment was protective for an initial colonization of the tracheobronchial tree (TBAS) and the lower respiratory tract (PSB) with Group I pathogens. Conversely, previous (short-term) antibiotic treatment was a risk factor for colonization of the tracheobronchial tree and the lower respiratory tract with Group II pathogens (Table 3).

A Glasgow Coma Score < 9 and previous (short-term) antibiotic treatment were associated with follow-up colonization with Group I pathogens of the upper airways. Treatment with barbiturates and previous (short-term) antibiotic treatment were associated with follow-up tracheobronchial colonization with Group I pathogens (Table 3).

Sequence of Colonization

Eleven of 24 patients (46%) with initial Group I colonization of the upper airways had follow-up tracheobronchial colonization with the same microorganism (including seven with concurrent initial tracheobronchial and two with gastric colonization). By the same token, six of the nine patients (67%) with initial Group II colonization of the upper airways had follow-up tracheobronchial colonization with the same microorganism (including three with concurrent initial tracheobronchial and five with gastric colonization).

Conversely, three of four patients (75%) initially colonized in the stomach with Group I pathogens had concurrent tracheobronchial colonization with the same microorganism. Six of 14 patients (43%) with initial Group II colonization of the stomach developed follow-up tracheobronchial colonization with the same microorganism. Of these, five had concurrent initial upper airway and two had tracheobronchial colonization with the same microorganism.

The inclusion of initial and follow-up upper airway and gastric colonization with Group I pathogens in a multivariate model revealed that initial upper airway colonization with Group I pathogens was an independent predictor of follow-up tracheobronchial colonization with Group I pathogens (odds ratio [OR], 9.9; 95% confidence interval [CI], 1.8 to 56.3; p < 0.01). Similarly, in a second multivariate model including initial and follow-up airway and gastric colonization with Group II pathogens, follow-up upper airway colonization with Group II pathogens was predictive for follow-up tracheobronchial colonization with these pathogens (OR, 23.9; 95% CI, 3.8 to 153.3; p < 0.001).

Non-PPMs and Candida spp.

The initial colonization rates with non-PPMs were highest in the upper airways (NS/PS: 44/46; 96%), followed by TBAS (28/41, 68%) and GJ (22/36, 61%), and lowest in PSB (16/36, 44%). At follow-up, rates were virtually unchanged (NS/PS: 41/43, 95%, p = 1.0; TBAS: 32/43, 74%, p = 0.53; GJ: 24/42, 57%, p = 0.72). With regard to Candida spp., initial colonization rates were very low in the upper airways (NS/PS: 4/46, 9%), TBAS (1/41, 2%), and PSB (1/36, 3%), but reached 25% (9/36) in GJ. At follow-up, all rates increased, but differences did not reach significance (NS/PS: 7/43, 16%, p = 0.28; TBAS: 6/43, 14%, p = 0.1; GJ: 17/42, 41%, p = 0.15) (Figure 3).

View larger version (22K): [in this window] [in a new window]   Figure 3.   Comparison of initial and follow-up colonization rates with non-PPMs and Candida spp. at different sites. NS = nasal swab; PS = pharyngeal swab; TBAS = tracheobronchial aspirate; PSB = protected specimen brush; GJ = gastric juice. Solid bars = initial colonization rate in %; dotted bars = follow-up colonization rate in %; p = NS for all comparisons.

Quantitative colonization patterns showed highest bacterial loads of non-PPMs in the pharynx (mean bacterial index, 12 ± 5.4) and the highest load of Candida spp. in the stomach (mean bacterial index, 1.5 ± 2.4). During Days 1 to 6, only the bacterial load of non-PPMs of the pharynx decreased significantly (12 ± 5.4 versus 5.3 ± 5.8, p = 0.02) (Figure 4).

View larger version (17K): [in this window] [in a new window]   Figure 4.   Quantitative colonization patterns in the nose, pharynx, trachea, and stomach. Log10 BI = mean sum of logarithms of colony counts, i.e., mean bacterial index (BI). Log10 FI = mean sum of logarithms of colony counts, i.e., mean fungal index (FI). Data are presented as mean ± SEM. Number of samples on sampling Days 1 to 6; see Figure 2. Dotted lines = non-potentially pathogenic microorganisms (non- PPMs); solid lines = Candida spp.

No predictor for colonization with non-PPMs or Candida spp. could be identified.

Incidence of Pneumonia

Pneumonia was clinically suspected in 21 (44%) patients (nine medical, 12 surgical), and microbiologically confirmed by PSB and/or TBAS in 10/21 (48%) patients and by blood cultures in one further patient. The presence of pneumonia was additionally confirmed at necropsy in one patient and by clinical response in seven patients. Thus, the incidence of confirmed pneumonia was 19/48 (40%). The mean duration of mechanical ventilation prior to the development of pneumonia was 5 ± 2 d. Early-onset pneumonia accounted for nine episodes (47%) (three medical, six surgical), and late-onset pneumonia for a further 10 episodes (53%) (six medical, four surgical). Prolonged antimicrobial pretreatment was present in 10 (53%) patients.

Risk Factors of Pneumonia

There was a trend for initial tracheobronchial colonization with Group I pathogens to predict early-onset pneumonia (OR, 4.1; 95% CI, 0.7 to 23.3; p = 0.09). Moreover, time-to-event analysis (Kaplan-Meier) revealed a significant association of initial tracheobronchial colonization with Group I pathogens with the probability to develop early-onset pneumonia (p = 0.02). No such association could be detected for the initial upper airway colonization with Group I pathogens (Figure 5).

View larger version (18K): [in this window] [in a new window]   Figure 5.   Kaplan-Meier plots for the probability to develop early-onset pneumonia in the presence of colonization with Group I pathogens in the upper airways and in the trachea. Solid lines = patients not colonized with Group I pathogens; dotted lines = patients colonized with Group I pathogens.

Follow-up tracheobronchial colonization with Group II pathogens (OR, 5.4; 95% CI, 1.0 to 29.6; p = 0.03), duration of mechanical ventilation (OR, 7.7; 95% CI, 1.4 to 41.6; p < 0.01) and prolonged antibiotic treatment (OR, 11.1; 95% CI, 2.0 to 61.9; p < 0.01) were predictors of late-onset pneumonia. A trend was also shown for initial tracheobronchial colonization with Group II pathogens (OR, 8.3; 95% CI, 0.9 to 75.8; p = 0.09). A multivariate analysis, including the three variables significantly associated with late-onset pneumonia, revealed that only prolonged antibiotic treatment was predictive of late-onset pneumonia (OR, 9.2; 95% CI, 1.7 to 51.3; p < 0.01).

Conversely, no association of non-PPMs and Candida spp. was found with early- or late-onset pneumonia, respectively.

Etiology of Pneumonia

The etiologic agents of early-onset pneumonia were most commonly Group I pathogens (four of five, 80%), whereas Group II pathogens were predominant in late-onset pneumonia (five of six, 83%). Four pneumonia episodes had polymicrobial etiology (two early-onset and two late-onset pneumonia episodes). All causative pathogens of pneumonia together with initial and follow-up colonization are shown in Table 4. Six of eight PPMs (75%) involved in early-onset pneumonia were already present in the upper airways, TBAS or PSB at the initial investigation and all eight (100%) during follow-up. Conversely, only two of eight PPMs (25%) involved in late-onset pneumonia were initially present, whereas seven of eight (88%) emerged during follow-up.

With regard to non-PPMs and Candida spp., the sequence of colonization and pneumonia was less evident. In early- onset pneumonia, non-PPMs in significant amounts (Streptococcus viridans) were present in only two of nine (22%) patients. These patients had previously been colonized at the upper airways (both) and in the tracheobronchial tree (one) with the same pathogen. In late-onset pneumonia, non-PPMs in significant amounts were present in four of ten (40%) patients (Streptococcus milleri, Streptococcus viridans [2], and Corynebacterium spp.). Of these, the latter three had been previously colonized with the same pathogen (all three in the upper airways, one each in the tracheobronchial tree and the stomach).

Outcome

Patients developing early-onset pneumonia had a comparable duration of mechanical ventilation (137 ± 63 versus 159 ± 116 h, p = 0.44) and length of ICU stay (11 ± 7 versus 9 ± 7 d, p = 0.41) as compared with those without. Results were similar when only survivors were included (duration of mechanical ventilation, 133 ± 73 versus 182 ± 109 h, p = 0.31; and length of ICU stay, 13 ± 7 versus 10 ± 6 d, p = 0.46).

Conversely, patients developing late-onset pneumonia had a significantly longer mean duration of mechanical ventilation (242 ± 95 versus 132 ± 100 h, p < 0.01) and longer ICU stay (14 ± 6 versus 8 ± 6 d, p = 0.02) as compared with those without. When only survivors were included, differences in duration of mechanical ventilation remained significant (261 ± 90 versus 148 ± 96 h, p < 0.01), whereas there remained only a corresponding nonsignificant trend for a longer ICU stay in patients with pneumonia (13 ± 5 versus 10 ± 7 d, p = 0.15).

Sixteen patients died (32%). There was no significant difference in mortality rates between patients with and without pneumonia (6/16, 38% versus 13/32, 41%; p = 0.83). Accordingly, mortality in the presence of early- and late-onset pneumonia was similar (3/9, 33% versus 3/10, 30%; p = 0.88). Pneumonia was the cause of death in only one patient with late-onset pneumonia.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main findings of the present study are the following. (1) Initial colonization rate in our study was very high (39/47, 83%). (2) Initial bacterial load with Group I pathogens in the upper airways and trachea was higher than bacterial load with Group II pathogens. A deflection point was observed (between Days 2 and 4) where the opposite occurred. The stomach was mainly colonized with Group II pathogens, with a slight increase of bacterial load during follow-up. (3) Previous (short-term) antibiotic treatment had a protective effect against initial and follow-up tracheobronchial and initial lower respiratory tract colonization with Group I pathogens. However, it was a risk factor for initial tracheobronchial and lower respiratory tract colonization with Group II pathogens. (4) Upper airway colonization independently predicted subsequent tracheobronchial colonization with the same group of pathogens. (5) Initial tracheobronchial colonization with Group I pathogens was associated with an increased probability of early-onset pneumonia, whereas prolonged antibiotic treatment independently predicted late-onset pneumonia.

Airway bacterial colonization is regarded as a common event in critically ill patients 2 or 3 d after the onset of mechanical ventilation (15), with airway colonization being directly related to the aspiration of pharyngeal and/or gastric contents (16). In our population, we found that upper and lower airway and/or gastric colonization was already initially present in 83% of patients as assessed by sampling within 24 h after intubation. This observation suggests that patients with head injury experience a compromise of local airway immune defense mechanisms very early during their illness, which allows microorganisms to adhere and persist at mucosal surfaces. Impaired levels of consciousness would additionally favor continuous aspiration and the subsequent development of pneumonia (17). This view is further supported by the fact that early colonization was found to be mainly due to community endogenous pathogens. Accordingly, it has been argued that tracheobronchial colonization with Staphylococcus aureus (18, 19) and Hemophilus influenzae (20) may result from direct inoculation during emergency intubation or endotracheal tube manipulation. In fact, in our study initial colonization of upper and lower airways mainly accounted for Group I pathogens, whereas Group II pathogens were only infrequently isolated. The reverse was true for gastric colonization. These colonization patterns were not different in surgical as compared with medical patients, indicating that head injury has similar consequences for local host defenses regardless of a history of trauma. The increase in colonization rates at all sites during follow-up was mainly due to colonization with Group II pathogens and developed within 48 h. This pattern of colonization accounts for the traditionally described events after initiation of mechanical ventilation.

The quantitative view on the evolution of colonization revealed two corresponding main colonization patterns. The first pattern demonstrable in the upper airways and the trachea showed a higher initial bacterial load with Group I pathogens, which steadily decreased during Days 3 to 6 of sampling; conversely, the bacterial load with Group II pathogens increased rapidly after Day 4, and at Day 6 of sampling always was significantly higher than the bacterial load with Group I pathogens. A second different pattern was evident in the stomach, which had a high initial and slightly increasing bacterial load with Group II pathogens but only a marginal initial and rapidly decreasing load with Group I pathogens. Two conditions obviously accounted for an important part of these dynamics: the presence of coma as well as of previous (short-term) and prolonged antibiotic treatment. A Glasgow Coma Score < 9 was associated with initial upper airway colonization with Group I pathogens, and a Glasgow Coma Score < 9 as well as treatment with barbiturates represented risk factors for follow-up colonization with these pathogens of the upper airways and the tracheobronchial tree, respectively. Treatment with barbiturates has also been shown by others to be dose-dependently associated with a higher incidence of colonization of the respiratory tract and pneumonia with Gram-positive bacteria in mechanically ventilated patients with cerebral edema (21). Although pentobarbitone is known to inhibit mucociliary clearance in sheep (22), also evidence for adverse effects of thiopentone on cellular (23) and humoral (24) immunological functions has been found. On the other hand, previous (short-term) antibiotic treatment was protective against an initial and follow-up tracheobronchial and initial lower airway colonization with Group I pathogens, whereas it represented a risk factor for initial colonization with Group II pathogens. Moreover, prolonged antibiotic treatment was an independent risk factor for late-onset pneumonia, which on the other hand was mainly due to Group II pathogens.

The qualitative and quantitative colonization patterns exposed so far support the concept of early- and late-onset pneumonia as entities with different pathogeneses and etiologies. From our data it is evident that the upper and lower airways are the main reservoir for early-onset pneumonia caused by Group I pathogens, which in turn account for the majority of early-onset pneumonia episodes. The sequence of initial Group I colonization of the upper airways with concomitant and/or subsequent tracheobronchial colonization with the same microorganism was demonstrable in 46% of patients with initial Group I colonization of the upper airways, and initial upper airway colonization with Group I pathogens was predictive for tracheobronchial follow-up colonization with these pathogens. Accordingly, antibiotic treatment covering Group I pathogens does protect against the colonization of the tracheobronchial tree with these pathogens. In line with these findings, we previously could show that short-term intravenous administration of high-dose cefuroxime (3 g in two repeated doses) has a protective effect against the development of early-onset pneumonia in patients with structural coma (5). However, beyond the fourth day of mechanical ventilation, colonization of all sites essentially accounted for Group II pathogens, and prolonged antibiotic treatment became potentially harmful. Both the upper airways as well as the stomach could be shown to initially harbor Group II pathogens concomitantly and/or subsequently identified in the tracheobronchial tree. These observations corroborate previous findings from our group, suggesting that both the upper airways as well as the stomach may represent an independent reservoir for late-onset pneumonia (10). However, we found that follow-up airway colonization with Group II pathogens was strongly associated with follow-up tracheobronchial colonization with these pathogens, suggesting the superior importance of the upper airway reservoir for late-onset tracheobronchial colonization and pneumonia. Other studies have suggested similar conclusions (25). However, the majority of our population was not ventilated for more than 1 wk, and thus these results must be interpreted with caution (32).

In our study, no independent risk factor for early-onset pneumonia could be determined. However, the probability of early-onset pneumonia was significantly increased in the presence of initial tracheobronchial colonization with Group I pathogens. We found that follow-up colonization of the tracheobronchial tree with Group II pathogens, the duration of mechanical ventilation, and prolonged antibiotic treatment were predictors of late-onset pneumonia, with the latter representing the only independent predictor. Corresponding findings have been reported in a previous study (33) and from other investigators (34, 35).

With regard to non-PPMs, colonization rates and quantitative colonization patterns did not follow those observed in PPMs. Likewise, the sequence of colonization and pneumonia was less evident for non-PPMs and Candida spp. The presence of these microorganisms did not predict early- or late-onset pneumonia, and given the very high initial and follow-up colonization rates of the upper airways, non-PPMs were only rarely found in pneumonia. There was also no evident association of non-PPMs to early- or late-onset pneumonia. These observations support the concept of excluding non-PPMs from the exposed pathogenetic framework of early- and late-onset pneumonia.

In conclusion, our study confirms the high incidence of VAP in patients with head injury. Patterns of colonization and pneumonia suggest that these patients suffer from an alteration of airway immune defense very early during their illness. The upper airways represent the most important reservoir of subsequent tracheobronchial colonization with Group I pathogens, which in turn is associated with early-onset pneumonia. Both the upper airways and the stomach may be independent reservoirs for tracheobronchial colonization with Group II pathogens and late-onset pneumonia. Preventive measures to reduce the incidence of early-onset pneumonia in this population may aim at an eradication of both upper and lower airway colonizers. In order to limit the adverse effect of a predisposition to colonization and pneumonia with Group II pathogens, the intravenous administration of antibiotics in future prophylaxis trials should be confined to less than 24 h.