Selective digestive decontamination has been widely used as a prophylactic regimen for ventilator-associated nosocomial pneumonia. The first to describe this complication, Stoutenbeek and colleagues [6] suggested that the best combination for preventing nosocomial pneumonia was the use of topical nonabsorbable antibiotics in the oropharynx and stomach together with systemic antibiotics. Most studies have shown a substantial decrease in the carriage of gram-negative bacilli of the upper and lower airways and also in the incidence of nosocomial pneumonia [7, 8], and a few studies have shown a substantial decrease in the overall mortality rate [9-11].

Several important considerations in most of the studies still make selective digestive decontamination a controversial issue. First, several studies were not randomized or used historical controls [6, 12-19]. Second, most of the randomized studies used only nonspecific methods to diagnose nosocomial pneumonia [9, 11, 20-28]. Finally, despite the apparent decrease in the incidence of nosocomial pneumonia, mortality did not change in most of the studies [12-28], including two recent randomized and double-blind studies [29, 30] of a large population sample of patients in an intensive care unit.

We did a randomized, double-blind study of selective digestive decontamination in a general population of patients requiring mechanical ventilation. The main end points of this study were to assess the effect of selective digestive decontamination in decreasing nosocomial pneumonia and mortality. Additional end points of this study were to determine the effect of selective digestive decontamination on the morbidity (length of stay and duration of mechanical ventilation) and the mortality rate.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

The study was done in the Respiratory Intensive Care Unit of the Hospital Clinic of Barcelona, Spain, a 1000-bed teaching hospital, during a period of 12 months. All mechanically ventilated patients admitted to the respiratory intensive care unit and expected to remain intubated for more than 3 days were included in the study. The only exclusion criterion was the presence of immunosuppression (human immunodeficiency virus [HIV] infection, HIV-related diseases, patients who received transplants, and patients treated with antineoplastic chemotherapy). Patients who were extubated or who died before receiving 72 hours of selective digestive decontamination or placebo were also excluded from the analysis.

Study Design

Patients were randomly allocated to either the selective digestive decontamination or the placebo group. The randomization was done using a computer-generated table, and the patients were enrolled consecutively. Severity of illness was evaluated by means of the Simplified Acute Physiologic Score after randomization. The authors of the study were blinded in the recovery of the results. The study ended after extubation or death of the patient in the intensive care unit.

Administration of Antibiotics

After samples for the bacteriologic assessment were obtained, antibiotics were administered for selective digestive decontamination. An aqueous suspension of 10 mL containing polymyxin E, 100 mg (Dumex; Dumex Limited, Denmark); tobramycin, 80 mg (Tobradistin; Dista SA, Madrid, Spain); and amphotericin B, 500 mg (Fungizona; Squibb Industria Farmaceutica SA, Madrid) was administered through a nasogastric tube to patients in the selective digestive decontamination group. Carboxymethyl-cellulose with pectin and with gelatin (0.5 mL, Orabase; Drogfesa, Mollet del Valles, Spain) containing polymyxin E, tobramycin, and amphotericin B, at 2% concentration, was applied four times a day.

In the placebo group, an aqueous suspension of Maxipro (Scientific Hospital Supplies Limited, Liverpool, United Kingdom) and Orabase, both colored with tartrazine, were administered through the nasogastric tube and in the oropharynx at the same dosage as for patients who received selective digestive decontamination.

Systemic Antibiotic and Stress Ulcer Prophylaxis

Patients were treated with 2 g of intravenous cefotaxime four times a day (Primafen, Hoechst Iberica SA, Barcelona, Spain) for the first 4 days of mechanical ventilation if they did not have infection on admission. Infected patients who were admitted to the intensive care unit received other parenteral antibiotics according to clinical decisions.

Prophylaxis for stress ulcers was done using 1 g of sucralfate every 4 hours (Urbal; Merck-Igoda SA, Mollet del Valles) through a nasogastric tube, except in patients with paralytic ileus or with upper gastrointestinal bleeding, who were treated with 50 mg of intravenous ranitidine, four times a day (Zantac; Glaxo SA-Allen Farmaceutica SA, Madrid).

Bacteriologic Assessment

Endotracheal aspirates, pharyngeal swabs, and gastric juice samples were obtained three times a week for quantitative cultures. Endotracheal aspirate samples were obtained by means of sterile tubes (Mocstrap; Productes Clinics, SA, La Llagosta, Barcelona). Samples obtained were diluted and homogenized in distilled water to 1/2 concentration using a vortex-style shaker (Reax 2000; Heidolph, Germany) and were rediluted in distilled water to 1/20 and 1/200 concentrations. Pharyngeal swabs were obtained using sterile swabs with Amies transport media (Eurotubo; Industrias Aulabor SA, Barcelona), were homogenized in 1 mL of distilled water, and were diluted to concentrations of 1/10, 1/10², and 1/10³.

Gastric juice samples were obtained by aspiration through a nasogastric tube using a sterile feeding syringe. The pH was determined in all the samples using paper indicators (Acilit, pH 0 to 6 and Spezialindikator, pH 6.5 to 10; Merck, Darmstadt, Germany). The samples were homogenized using a vortex-style shaker and were diluted in distilled water to concentrations of 1/10 and 1/100.

All samples were plated on the following agar media: blood; chocolate; McConkey-2; buffered, charcoal, and yeast extract (BCYEa); Sabouraud-dextrose; Sabouraud with nalidixic acid; and blood with nalidixic acid. If negative, the plates were discarded after 5 days of testing for aerobic bacteria, after 10 days of testing for Legionella and anaerobic bacteria, and after 4 weeks of testing for fungi. If positive, counts of colony-forming units per milliliter and identification using standard methods [31] were done for the microorganisms.

Definitions

Potentially pathogenic microorganisms were defined [32] as those causing infection in a person with impaired defense mechanisms. They can be classified into "community" microorganisms, which cause infections in previously healthy persons with intact carriage defense, and nosocomial microorganisms, which cause infections in persons with impaired carriage defense.

Colonization was defined as the isolation of the same strain of a potentially pathogenic microorganism from at least two consecutive surveillance samples in any concentration. The clinical diagnosis of pneumonia was based on the presence of all of the following criteria: new or progressive pulmonary radiologic infiltrate or both for 48 hours or more, purulent tracheal secretions, temperature of 38.5 °C or more, and leukocytosis ( 12 10⁹/L) or leukopenia ( 4 10⁹/L). The diagnosis of pneumonia was confirmed by the isolation of a potentially pathogenic microorganism in a protected specimen brush sample in concentrations of 10³ CFU/mL or more or in a bronchoalveolar lavage sampling in concentrations of 10⁴ CFU/mL or more [33]. We defined definite pneumonia when all the clinical criteria and one bacteriologic criterion were present or by the presence of histologic signs of pneumonia at autopsy. Probable pneumonia was defined when only clinical criteria were present.

Primary endogenous pneumonia was diagnosed when pneumonia developed within the first 4 days of mechanical ventilation and when etiologic microorganisms were isolated previously or concomitantly in pharyngeal swabs or in gastric juice. Secondary endogenous pneumonia was pneumonia that developed after the fourth day of mechanical ventilation. Exogenous pneumonia was diagnosed when the etiologic microorganism was not isolated in pharyngeal swabs or in gastric juice before the development of pneumonia. Community flora was defined as the isolation of normal buccal flora (Neisseria species, Streptococcus viridans, among others), Streptococcus pneumoniae, or Haemophilus influenzae.

A catheter-related infection was diagnosed when inflammatory signs occurred in a catheterized blood vessel together with a temperature of 38.3 °C or more, irrespective of the isolation of a potentially pathogenic microorganism in the culture of the removed catheter. Likewise, this diagnosis was considered if the fever improved within 12 hours after removing the catheter. A urinary tract infection was diagnosed after fresh-voided catheter urine containing five or more leukocytes per high-power light-microscopic field were identified and a potentially pathogenic microorganism was isolated in urine culture in concentrations of 10⁵ CFU/mL or more.

A wound infection was diagnosed if purulent secretions from wounds occurred with signs of inflammation and the isolation of a potentially pathogenic microorganism in concentrations of 10⁵ CFU/mL or more from the purulent wound secretions. Septicemia was diagnosed if clinical signs of systemic infection occurred, such as fever, leukocytosis, increased percentage of band forms, and metabolic acidosis, combined with a positive blood culture.

Multiple organ system failure was defined as three or more organ systems failing for more than 2 consecutive days. Infection-related death was defined as death during a clinically confirmed infection. The autopsy, when done, needed to have active signs of infection at any site.

Statistical Analysis

The sample size (80 patients) was calculated to allow the detection of a 25% decrease in nosocomial pneumonia (from an expected rate of 35% that decreased to 10%, which was derived from previous studies in the placebo group) with an error of less than 0.05 and with a ß error of 0.20. Consequently, only dramatic decreases in the rate of nosocomial pneumonia would have been detected with an 80% certainty. The statistical analysis was done using the SPS/PC+ 4.0 (Statistic Package for Social Sciences) for the IBM PC/XT/AT and PS/2 (SPS Inc., Chicago, Illinois). Qualitative variables were compared using the homogeneity test of chi-square and the Fisher exact test when chi-square was not applicable. Quantitative variables were compared using the unpaired t-test and the Mann-Whitney nonparametric test when the t-test was not applicable.

Results

Top Methods Results Discussion Author & Article Info References

One hundred and one patients were enrolled in the study (51 in the selective digestive decontamination group and 50 in the placebo group). Eighty remained eligible for the trial (39 in the selective digestive decontamination group and 41 in the placebo group). The underlying conditions of the patients and the main clinical characteristics of the population studied were similar for the decontamination and placebo groups: age, 62 ±18 years compared with 59 ±20 years; simplified acute physiologic score, 13 ±4 compared with 12 ±4; types of patients studied: medical, 24 compared with 28; postoperative, 6 compared with 4; and trauma, 9 compared with 9.

Bronchial colonization by gram-negative bacilli was lower in patients receiving selective digestive decontamination (12 of 39 [31%]) compared with those receiving placebo (31 of 40 [78%]; odds ratio, 7.75; 95% CI, 2.5 to 24.4). The most common gram-negative bacillus isolated was Pseudomonas aeruginosa. The rate of colonization by this microorganism was lower in the selective digestive decontamination group (9 of 39 [23%]) than in the placebo group (25 of 40 [63%]; odds ratio, 5.6; CI, 1.9 to 16.9). The colonization rate by methicillin-resistant Staphylococcus aureus was similar in the selective digestive decontamination group and placebo groups: 14 of 39 patients (36%) compared with 12 of 40 patients (30%; odds ratio, 0.75; CI, 0.27 to 2.2). Colonization by Candida species was lower in the selective digestive decontamination group (8 of 39 patients [21%]) than in the placebo group (16 of 40 patients [40%]; odds ratio, 2.5; CI, 0.9 to 8). Differences in colonization of the oropharynx and gastric cavities between both groups showed results similar to those found in upper airways (Table 1).

No significant differences (P > 0.2) were noted between the two groups in incidence of nosocomial pneumonia: 7 of 39 patients in the selective digestive decontamination group (18%) and 10 of 41 patients in the placebo group (24%) (odds ratio, 1.44; CI, 0.44 to 4.99). The etiologic agents of these episodes of pneumonia, identified mainly on the basis of quantitative cultures of protected specimen brush samples, are summarized in Table 2. Definite pneumonia was diagnosed in 6 and 9 patients in the selective digestive decontamination and placebo groups, respectively. Two patients were considered to have probable pneumonia, one in each group. From all the pneumonias diagnosed, only one in the placebo group was considered as primary endogenous compared with none in the selective digestive decontamination group. Four secondary endogenous pneumonias were diagnosed in the selective digestive decontamination group compared with seven in the placebo group. Two pneumonias were considered as potentially exogenous in the selective digestive decontamination group compared with none in the placebo group. As for gram-negative bacilli pneumonia, two patients (patients 6 and 40, Table 2) in the selective digestive decontamination group showed absence of previous or concomitant pharyngeal or gastric colonization. These pneumonias were considered of probable exogenous origin. In the placebo group, four patients (patients 11, 15, 16, and 18, Table 2) showed colonization of the oropharynx or gastric juice or both by the same microorganism isolated from protected specimen brush samples. These pneumonias were considered as secondary endogenous. Methicillin-resistant Staphylococcus aureus pneumonias in the two groups were of secondary endogenous origin. Two additional patients who received placebo and selective digestive decontamination, respectively, had primary endogenous pneumonia that was caused by Candida tropicalis and Enterococcus faecalis. Finally, in the placebo group, two patients had secondary endogenous pneumonia that was caused by "community" flora.

Four patients in the selective digestive decontamination group and six in the placebo group developed catheter-related infections. These four patients in the digestive decontamination group had infections that were caused by methicillin-resistant Staphylococcus aureus, Staphylococcus epidermidis, and P. aeruginosa; no microorganism was isolated in the remaining patient. The six patients in the placebo group had infections that were caused by methicillin-resistant Staphylococcus aureus in two patients, Staphylococcus epidermidis, P. aeruginosa, and Candida albicans; no microorganism was isolated in the remaining case. The rate of bacteremia was similar in both groups; two cases occurred in each. The two patients who had infections in the digestive decontamination group had catheter-related sepsis caused by P. aeruginosa and Staphylococcus epidermidis, respectively. The two patients in the placebo group had catheter-related sepsis and pneumonia, both caused by methicillin-resistant Staphylococcus aureus.

The onset of nosocomial pneumonia and the duration of mechanical ventilation and stay in the intensive care unit did not change when both groups were compared. The number of days of mechanical ventilation at the onset of nosocomial pneumonia was 10.4 ±7.2 days for the selective digestive decontamination group and 8.9 ±3.2 days for the placebo group. The duration of mechanical ventilation was 13.5 ±9.3 days for the digestive decontamination group and 12.6 ±7.4 days for the placebo group. The duration of time in the intensive care unit was 15.3 ±10 days for the digestive decontamination group and 14.3 ±7.3 days for the placebo group.

The crude mortality rate was similar in the two groups: twelve of 39 patients (31%) in the selective digestive decontamination group and 11 of 41 patients (27%) in the placebo group died (odds ratio, 0.82; CI, 0.28 to 2.41). Six patients in the digestive decontamination group died of multiple organ system failure, 2 patients of head trauma, 2 of nosocomial pneumonia, 1 of nosocomial pneumonia with multiple organ system failure, and 1 of cerebral hemorrhage. In the placebo group, 3 patients died of head trauma, 2 of nosocomial pneumonia, 2 of multiple organ system failure, 1 of nosocomial pneumonia with multiple organ system failure, 1 of methicillin-resistant Staphylococcus aureus sepsis, 1 of the adult respiratory distress syndrome, and 1 of nosocomial pneumonia in the context of the adult respiratory distress syndrome. The incidence of attributable mortality because of pneumonia was lower in the digestive decontamination group than in the placebo group (2 compared with 4 patients), but these differences were not significant. We observed a higher rate of multiple organ failure in the digestive decontamination group (7 patients compared with 3 patients receiving placebo), yet these differences were not significant.

Discussion

Top Methods Results Discussion Author & Article Info References

The most important finding of our study was that no significant differences occurred in the incidence of nosocomial pneumonia and in the incidence of other infections when comparing patients who received selective digestive decontamination with the placebo group. The length of stay, the duration of mechanical ventilation, and the overall mortality rate remained unaltered when both groups of patients were compared. A statistically significant decrease occurred in the colonization rate by gram-negative bacilli and yeast in the bronchial tree, the oropharynx, and the stomach after the use of digestive decontamination. These results concur with two recently published randomized, double-blind studies [29, 30] that included a large number of general mechanically ventilation patients. In the first study, Gastinne and colleagues [29] analyzed a group of 445 patients receiving mechanical ventilation; they did not find statistical differences between the selective digestive decontamination group and the placebo group for either the incidence of pneumonia (12% compared with 15%) or the mortality rate (34% compared with 30%). In the second study, Hammond and colleagues [30] analyzed 239 patients receiving mechanical ventilation and found similar results; thus, the incidence of infection was 26% compared with 34%, and the mortality rate was 18% compared with 17%.

The incidence of nosocomial pneumonia among mechanically ventilated patients is still controversial and ranges between 9% [34] and 70% [35], with an average incidence of 20%. The incidence of ventilator-associated pneumonia is greater in medical patients (33%) [36] compared with patients (17%) who have had surgery (postoperative patients) [37]. The discrepancies are attributable to the types of patients studied and the lack of uniformity in the diagnostic methods used. The results of our study showed a similar incidence of ventilator-associated nosocomial pneumonia in patients receiving either selective digestive decontamination or placebo (18% compared with 24%, respectively). Because all patients received systemic antibiotics within the first 4 days or during several days to treat initial infections, we cannot infer from our study whether or not selective digestive decontamination, in addition to other systemic antibiotics, can prevent primary endogenous pneumonia. Instead, we can only conclude that digestive decontamination did not prevent secondary endogenous pneumonias in our patients.

Our results about the effects of selective digestive decontamination on the incidence of pneumonia do not agree with results of most of the randomized controlled trials that have shown statistically significant decreases in the incidence of nosocomial pneumonia. These differences in results may be because none of the previous studies [6, 27, 30] used techniques with high specificity, such as protected specimen brush samples or bronchoalveolar lavage specimens, for the diagnosis of nosocomial pneumonia. Further, the diagnosis of nosocomial pneumonia in these previous studies was based on the presence of clinical and radiographic signs with or without positive cultures of tracheal secretions. Several studies [35, 38, 39] have shown the inaccuracy of the clinical variables in diagnosing ventilator-associated nosocomial pneumonia, particularly in patients with the adult respiratory distress syndrome. In these studies, the presence of clinical signs of respiratory infection has shown a high incidence of false-positive and false-negative results. Further, using endotracheal aspirates is not specific for diagnosing nosocomial pneumonia; it is widely accepted that the most sensitive and specific techniques to diagnose ventilator-associated nosocomial pneumonia are bronchoscopic procedures, such as culturing protected specimen brush samples or culturing bronchoalveolar lavage specimens [33, 34, 36, 40-43].

These two important methodologic considerations could have biased the results of several trials examining selective digestive decontamination. The only digestive decontamination trial using systematically protected specimen brush samples to diagnose nosocomial pneumonia was done by Brun-Buisson and colleagues [44]. In that study, differences were not detected in the incidence of nosocomial pneumonia or in the mortality rate. A third bias in these studies is that there was no differentiation between exogenous or endogenous respiratory infections. However, we found that some pneumonias were exogenous in origin; for future studies, we suggest that this should be considered.

The incidence of nosocomial pneumonia based on microbiologic criteria may be underestimated. Gastinne and colleagues [45] detected lower tobramycin and amphotericin B levels in distal airway than in tracheal secretions of patients on mechanical ventilation who received intestinal and oropharyngeal decontamination. Antibiotic levels were greater than the minimal inhibitory concentration for gram-negative bacilli. The presence of antibiotics in upper airway secretions sterilizes the cultures of endotracheal aspirates and, consequently, falsely diminishes the real incidence of pneumonia. In our study, because protected specimen brush samples were used, the effect of digestive decontamination on microbiologic cultures was minimized. Thus, we believe that our results reflect the actual incidence of nosocomial pneumonia in the two groups of patients.

The mortality rate was similar in both groups (31% in the selective digestive decontamination group compared with 27% in the placebo group), although the number of pneumonia-related deaths was greater in the placebo group (two patients compared with four patients, respectively). The small number of patients who died limits the inferences that can be drawn from the study, but the absence of any effects of digestive decontamination on mortality is consistent with previous studies [6, 11-30].

When considering the use of selective digestive decontamination, we believe that it is better initially to use less complicated and cheaper tools for preventing nosocomial pneumonia. Recently, it has been suggested that the use of sucralfate (Carafate) instead of histamine-2 blockers or antacids [47], the body position [48], hand washing [49], and the use of double-lumen endotracheal tubes [50] are simple and effective measures for preventing nosocomial pneumonia.

Our results suggest that selective digestive decontamination does not decrease secondary pulmonary infections in a general population of patients receiving mechanical ventilation. Although new multicenter studies with a large number of patients seem necessary to establish the precise role of digestive decontamination, it would be better to focus investigations toward the use of a less complicated and cheaper method to prevent nosocomial pneumonia. We recommend that all these prevention studies should use specific diagnostic techniques for identifying pneumonia, such as culturing protected specimen brush samples or bronchoalveolar lavage samples. We admit that only dramatic decreases in the rate of nosocomial pneumonia would have been detected with reasonable statistical power (80% certainty) in our study. Smaller decreases, even if statistically significant, may not be clinically meaningful because of the added cost and the increased risk for selecting microbial strains.

Presented in part at the 1992 International Conference of the American Lung Association/American Thoracic Society in Miami, Florida.