Selective decontamination of the digestive tract: a meta-analysis
ACP J Club. 1994 May-June;120:67. doi:10.7326/ACPJC-1994-120-3-067
Selective Decontamination of the Digestive Tract Trialists' Collaborative Group. Meta-analysis of randomised controlled trials of selective decontamination of the digestive tract. BMJ. 1993 Aug 28;307:525-32.
To determine the efficacy of selective decontamination of the digestive tract (SDD) in preventing respiratory tract infection and death in patients in the intensive care unit (ICU).
MEDLINE was searched (January 1984 to June 1992) using the terms intensive care units; critical care; antibiotics combined with therapeutic use and administration and dosage; and respiratory tract infections combined with prevention and control. Investigators who had published on the topic were contacted and proceedings of meetings were examined to identify unpublished studies.
Studies were selected if they were randomized controlled trials of SDD for the prevention of respiratory tract infections and death in ICU patients. Patients had an expected ICU stay of ≥ 3 days or mechanical ventilation for ≥ 48 hours. 22 relevant trials were identified.
For each study the number of patients, method of randomization, and proportions of medical, surgical, and trauma patients were recorded. Respiratory tract infections, deaths, and the number of patients excluded for an intention-to-treat analysis were determined for each study. To assess the effect of selective decontamination (SDD) on respiratory tract infections and death, studies were subgrouped according to the method of randomization, blinding, antibiotic regimen (topical vs. systemic plus topical), and diagnostic approach to respiratory tract infection. The numbers of patients that needed to be treated to prevent 1 respiratory tract infection and 1 death were calculated.
22 studies evaluating 4142 patients (mean age range 43 to 65 y) met the inclusion criteria. SDD reduced respiratory tract infections overall (common odds ratio 0.37, 95% CI 0.31 to 0.43). The number of patients with a baseline risk of 30% that would need to be treated with SDD to prevent 1 respiratory infection was 6 (CI 5 to 7). There was no difference between treated patients and controls for overall mortality (common odds ratio 0.90, CI 0.79 to 1.04). Subgroup analysis of combined intravenous, topical, and oral antibiotics showed a reduction in mortality (common odds ratio 0.80, CI 0.67 to 0.97). Based on a risk for death in untreated patients of 30%, 23 patients would need to be treated to prevent 1 death (CI 13 to 161).
Selective decontamination of the digestive tract was effective in reducing respiratory tract infections in critically ill patients but did not decrease mortality.
Source of funding: Not stated.
For article reprint: The Secretariat, Selective Decontamination of the Digestive Tract Trialists' Collaborative Group, Laboratorio di Epidemiologia Clinica, Istituto Mario Negri, Via Eritrea 62,20157 Milan, Italy. Fax 39-2-3320-0231.
Nosocomial infection remains a major problem in the ICU because it may be associated with an increased incidence of organ failure and death. Evidence suggests that the organisms responsible are those that colonize the gastrointestinal and respiratory tracts of susceptible hosts. SDD is a technique designed to influence this colonization, suppressing the growth of potentially pathogenic aerobic gram-negative bacteria while allowing the continued presence of normal anaerobic bacteria in the gut.
Although many groups have reported a decrease in respiratory infections, few have reported any decrease in mortality. Several investigators concluded that this failure to show a decrease in mortality may be caused by study design. The meta-analyses by Heyland and colleagues and the Selective Decontamination of the Digestive Tract Trialists' Collaborative Group (TCG) attempted to address this problem as well as to summarize quantitatively all studies to date. Both meta-analyses used similar rigorous methods to retrieve, review, and analyze previous studies. Additionally, the TCG used intention-to-treat analysis by asking for original data from all trialists. The TCG analyzed 22 randomized clinical trials, whereas Heyland and colleagues analyzed 25. 20 were examined by both groups.
Not surprisingly, perhaps, the studies had similar results. A significant difference in respiratory infection rates existed between treatment and control groups, but neither study showed any convincing difference in mortality. Thus, neither paper has produced results that differ importantly from those of the original studies or of previous reviews. Both papers conclude that further evaluation is necessary to determine whether mortality is affected, in part because, even with these large meta-analyses, the sample size may have been insufficient. Heyland further recommends that widespread use of SDD is not warranted at the present time.
Answers to the question of whether SDD reduces mortality, therefore, continue to remain elusive. Aside from sample-size considerations, both meta-analyses are limited by the amount of information available from the original randomized controlled trials. Of most interest is the potential to decrease mortality in particular patient populations, such as young adults with trauma. Neither meta-analysis had information on the outcome of individual patients within the original studies and, therefore, could not pool outcomes for these specific patient populations.
If a clearly shown decrease in mortality is required before SDD is routinely used, then we must await the results of future research. If, on the other hand, a decrease in respiratory infection is of importance, then clinicians may wish to use SDD now. This use, however, must be with the caveat that the presumed benefit of reduced morbidity and, therefore, reduced resource utilization has not been shown. Indeed, current evidence, although based on crude proxies of resource utilization, suggests little or no benefit (1).
Both meta-analyses have suggested that future evaluation of SDD must adhere to a scrupulous clinical trial design, include adequate sample size, and address a priori the specific ICU populations to be studied. If a heterogeneous population is studied, information on individual outcomes, as shown by the TCG, should be retained to allow improved meta-analysis.
Further work also is required on the mechanism of action of SDD. Both reviews were based on the underlying premise of the randomized controlled trials that SDD reduced mortality via a reduction in respiratory infection. The gut may play a central role in the initiation and maintenance of the systemic inflammation seen in critical illness, regardless of nosocomial infection. Recent work suggests SDD may influence that inflammation by reducing gut translocation (2). As we come to better understand the role of the gut in critical illness, we may be able to better determine the value of SDD.
Derek C. Angus, MB, ChB, MPH
University of Pittsburgh School of Medicine Pittsburgh, Pennsylvania, USA