Velmahos GC, Toutouzas KG, Sarkisyan G, Chan LS, Jindal A, Karaiskakis M, Katkhouda N, Berne TV, Demetriades D. Severe Trauma Is Not an Excuse for Prolonged Antibiotic Prophylaxis. Arch Surg. 2002;137(5):537-542. doi:10.1001/archsurg.137.5.537
For critically injured patients, a limited course of antibiotics is as effective as a prolonged course in preventing sepsis and organ failures.
Prospective nonrandomized study.
Surgical intensive care unit (SICU) of an academic hospital with a level I trauma center.
A population of 250 trauma patients who required an operation and SICU stay of 3 days or more received antibiotic prophylaxis by 1 antibiotic for 24 hours (SHORT group, n = 133) or 1 or more antibiotics for more than 24 hours (LONG group, n = 117).
Main Outcome Measures
Twenty-two outcome variables, including 9 conventional outcomes (eg, sepsis, septic shock, and organ failure) and 13 objective outcomes (days with temperature >38.5°C, days with white blood cell count >14.0 ×103/µL, positive cultures, cultures with antibiotic-resistant bacteria, SICU and hospital stay, and death).
The LONG group included more patients with orthopedic injuries (60 patients [51%] vs 52 [39%], P = .05) and orthopedic operations (47 patients [40%] vs 30 [23%], P = .003) than did the SHORT group. No other difference was identified in compared characteristics between the 2 groups. There was no difference in any of the examined outcomes except for a higher incidence of resistant infections in the LONG group compared with the SHORT group (59 patients [50%] vs 47 [35%], P = .02). Patients with resistant infections stayed in the hospital longer (mean ± SD, 33 ± 18 vs 15 ± 11 days, P<.001) and had a higher mortality rate (13% vs 1%, P<.001) compared with patients without resistant infections. Prolonged prophylaxis by multiple antibiotics was an independent risk factor of resistant infection (odds ratio, 2.13, 95% confidence interval, 1.22-3.74; P = .008).
The prophylactic administration of more than 1 antibiotic for more than 24 hours following severe trauma does not offer additional protection against sepsis, organ failure, and death, but increases the probability of antibiotic-resistant infections.
ACCORDING TO Webster's Dictionary,1 the word prophylaxis—based on the Greek word phylaxis, which means a watching—is interpreted as "prevention of or protection from disease." Antibiotic prophylaxis is currently being used successfully for prevention of infectious morbidity following elective operations.2,3 Published evidence2,4,5
describes the following necessary elements of success: (1) administration before the surgical insult begins, (2) coverage targeted to organisms specific to the operation, and (3) short duration of antibiotic administration. Although these rules fit perfectly in elective surgery, their application in trauma cases is questionable. Antibiotics can never be given before the injury and unavoidable contamination occur and therefore are presumptive rather than purely prophylactic. There is usually more than 1 injured organ with different bacteriologic floras, precluding the selection of regional flora-targeting narrow-spectrum antibiotics. Finally, the effectiveness of a short course, although potentially adequate for mild or moderate injuries, may be insufficient for severe trauma. Only a few of the existing studies on posttraumatic antibiotic prophylaxis have analyzed high-risk groups.6,7 Many others include predominantly or exclusively patients without critical injuries.8- 13
In our trauma center, some trauma surgeons consistently treat severely injured patients with limited antibiotic prophylaxis, while others administer a longer course. Our orthopedic consultants prescribe routinely 2 antibiotics for at least 3 days for multitrauma patients with open fractures. The objective of our study was to compare the mortality, infectious morbidity, and hospital stay between severely injured patients who received antibiotic prophylaxis for 24 hours with a single agent and patients who received longer courses of prophylaxis with 1 or more antibiotics.
For 29 months (January 1, 1999, to May 31, 2001), critically injured patients who required care in our 16-bed surgical intensive care unit (SICU) for 3 or more days were included prospectively in the study. Patients were excluded if they had isolated head or spinal cord injuries, did not undergo an operation, did not receive prophylactic antibiotics, or were admitted to the SICU following deterioration in a regular ward. The SICU is managed 24 hours a day by a group of 6 qualified trauma surgeons and 1 intensivist.
Antibiotic prophylaxis was given according to the individual trauma surgeon's preference but always before the skin incision. Two groups of patients were identified. The SHORT group received only 1 broad-spectrum antibiotic for 24 hours. This antibiotic was cefoxitin sodium or ampicillin sodium/sulbactam sodium. The LONG group received 1 or more antibiotics for more than 24 hours, typically for 3 to 5 days. Various antibiotics were used; the most frequent ones were a combination of piperacillin sodium sterile and tazobactam sodium, a second- or third-generation cephalosporin, or the combination of ampicillin, gentamicin sulfate, and metronidazole. Because of the many different regimens used, we did not attempt to create subgroups according to type, number, or duration of antibiotics.
We collected data on patient demographics, mechanism of trauma, injury severity as expressed by the Injury Severity Score (ISS) and the 6 Abbreviated Injury Scores, physiological response, blood transfusion, and operative intervention. Recognizing the limitations of the definitions of commonly used morbidities, we decided to use a variety of outcome variables. The first set of such variables included conventional outcomes, which are typically defined arbitrarily on the basis of consensus conferences or expert opinions. These 9 outcomes were sepsis, septic shock, renal failure, respiratory failure, cardiac failure, liver failure, any organ failure, multiple organ failure syndrome, and adult respiratory distress syndrome.14 The second set of variables included objective outcomes and did not require arbitrary definitions. These 13 outcomes were number of days with temperature higher than 38.5°C, number of days with white blood cell count higher than 14.0 ×103/µL, number of days with platelet count lower than 50.0 ×103/µL, number of days with total serum bilirubin levels higher than 6 mg/dL (103 µmol/L), number of days with serum creatinine levels higher than 2.5 mg/dL (221 µmol/L), number of days with a partial oxygen arterial tension to fractional inspired oxygen ratio lower than 200, number of days with mean arterial blood pressure lower than 70 mm Hg, number of patients with bacterial growth in blood cultures (positive blood cultures), number of patients with bacterial growth in any culture (positive cultures), number of patients with antibiotic-resistant bacteria on any culture, length of SICU stay, length of hospital stay, and death. All patients with cultures growing antibiotic-resistant bacteria were treated and therefore considered to have resistant infections.
The SHORT group was compared with the LONG group on the basis of the 22 outcomes. Categorical variables were compared by the χ2 test or, in the case of small numbers, the Fisher exact test. Continuous variables were compared by the t test or Wilcoxon 2-sample test. Multiple groups were compared by analysis of variance. Independent risk factors of the 4 selected outcomes, sepsis, organ failure, resistant infection (as defined by the presence of antibiotic-resistant bacteria in any culture), and death, were identified by stepwise logistic regression. Adjusted odds ratios and 95% confidence intervals were calculated for all independent risk factors. Specific attention was drawn in evaluating whether the method of antibiotic prophylaxis (ie, the group allocation) was an independent risk factor for any of the selected outcomes. Statistical significance was considered at the P<.05 level. Data are presented as mean ± SD unless otherwise specified.
A population of 250 patients with a mean age of 33 ± 14 years were included, 133 (53%) in the SHORT group and 117 (47%) in the LONG group. The responsible mechanism of injury was penetrating trauma in 135 patients (54%) and blunt trauma in 115 (46%). The mean ISS was 22 ± 10, with 76% of the patients having an ISS of 16 or more and 41% an ISS of 25 or more. One hundred eighty-four patients (74%) underwent an abdominal or thoracic operation; 77 (31%), an orthopedic operation; and 14 (6%), a craniotomy. A hollow visceral injury was found in 57 patients (23%), and, more specifically, a colon injury in 38 (15%). An average blood transfusion of 6 ± 7 units was required in the first 24 hours, with two thirds of the patients requiring 3 or more units. Sixteen patients (6%) died, 108 (43%) developed sepsis, and 102 (41%) had organ failures. The average SICU and hospital lengths of stay for survivors were 13 ± 13 and 22 ± 17 days, respectively.
The only statistical differences found on comparison of different characteristics of the 2 groups were in the incidence of extremity-pelvic trauma, extremity Abbreviated Injury Score of 3 or more, and orthopedic operations (Table 1). Similarly, no statistically significant differences existed between the 2 groups with regard to the 9 conventional outcomes (Table 2) and to 12 of the 13 objective outcomes (Table 3). The only difference was found in the incidence of cultures with antibiotic-resistant bacteria (35% in the SHORT vs 50% in the LONG group, P = .02). This difference was based mainly on the cultures from bronchial secretions (Table 4). Although specific complications, such as abscess, wound sepsis, and fistula, were not outcome points of this study, we classified the source of sepsis on the basis of the patient's cultures and clinical history. No difference was found between the 2 groups in this comparison (Table 5).
Stepwise logistic regression identified independent risk factors for the 4 selected variables, as shown in Table 6. Notably, being in the LONG group was an independent risk factor for the development of resistant infections. Compared with patients without resistant infections, patients with resistant infections stayed longer in the SICU and the hospital (7 ± 5 vs 20 ± 15 days [P<.001] and 15 ± 11 vs 33 ± 18 days [P<.001], respectively) and had a higher mortality rate (1% vs 13% [P<.001]).
Various subgroup analyses were performed although limited by small sample sizes.
Of the 117 patients in the LONG group, 65 received 1 antibiotic; 41, 2; and 11, more than 2. On comparison of these subgroups with the SHORT group, sepsis was recorded in 53 patients (40%) in the SHORT group vs 31 patients (48%) in the LONG group who received 1 antibiotic, 16 (39%) who received 2 antibiotics, and 8 (73%) who received more than 2 antibiotics (P = .15). Organ failure was recorded in 48 patients (36%), 26 (40%), 22 (54%), and 6 (55%), respectively (P = .18). Finally, 6 patients (5%), 6 (9%), 4 (10%), and 0, respectively, died (P = .36).
Forty-five patients in the SHORT group and 46 in the LONG group had blunt trauma and an ISS of 16 or more, both of which were found to be independent risk factors for adverse outcomes. Sepsis occurred in 27 of these SHORT group patients (60%) compared with 25 LONG group patients (54%) (P = .67); organ failures, in 30 patients (67%) compared with 29 (63%) (P = .72); and death, in 5 patients (11%) compared with 7 (15%) (P = .76).
Among 41 patients of the LONG group, 29 received 2 antibiotics for 3 days for orthopedic injuries and 12 for nonorthopedic injuries. Resistant infections developed in 20 (69%) of those with orthopedic injuries compared with 4 (33%) with nonorthopedic injuries (P = .045). There was no difference in the occurrence of sepsis, organ failure, or death.
Throughout the world, the practices of antibiotic prophylaxis after trauma follow no standard scientific principles. Surgeons administer a variety of prophylactic antibiotics for a variety of periods, despite evidence suggesting that such policies may be wrong or even harmful.15,16 In a survey of 342 members of the Association for the Surgery of Trauma, 64% stated that they would prescribe antibiotics for more than 24 hours for an isolated colon injury.17 Despite the existence of hospital-specific guidelines for antibiotic administration in critically injured patients, physicians are not likely to follow these guidelines.18,19 Multiple reasons are cited for the tendency of surgeons to favor long prophylactic courses of multiple antibiotics, including the inability to distinguish between infection and contamination or inflammation and the dogmatic refusal to understand the risks of indiscriminate antibiotic administration.15,20
In the presence of injury to multiple body areas, the variety of pathogens that may require antibiotic coverage increases the confusion and results in a "shotgun" approach to prophylaxis. Calls for scientific rationalization and encouragement for additional research in this area are indicative of the importance of the problem.21,22
The existing trauma literature8- 13 supports overwhelmingly the use of limited antibiotic prophylaxis. However, the main shortcoming of these studies is the lack of inclusion of enough high-risk patients. It is reasonable to assume that patients at low risk for sepsis, organ failure, and death will usually do well with or without major antibiotic prophylaxis. It is the patients with severe trauma who need drastic measures to curtail the risk of death from sepsis and sepsis-related organ failures. To our knowledge, there is only one study that specifically addresses high-risk patients6 and one more that subanalyzes a group of patients at high risk for abdominal sepsis.7
Both articles are limited to penetrating abdominal trauma and conclude that a short course of prophylaxis is as effective as longer courses.
In our study, we evaluated an unselected group of severely injured patients and included a variety of injuries from blunt and penetrating trauma because it is not practical to make different antibiotic policies for different injuries. The more complex the policies are, the less probable their implementation becomes. Physicians would have to remember numerous guidelines and algorithms. It is more realistic to make a universal policy on antibiotic prophylaxis for all critically injured patients, allowing for variations under individual and well-specified conditions.
Although our study was not randomized, a strong selection bias is not likely because the individual trauma surgeons were consistent with their behavior toward severe trauma, ie, either all their critically injured patients were given extended antibiotic prophylaxis or all received limited prophylaxis. The comparison of the 2 groups confirms this argument, as shown by the lack of differences in multiple characteristics, except those related to orthopedic injuries and those reflecting the policy of our orthopedic department.
We selected a variety of outcomes to uncover any potential effect of our antibiotic interventions, even if that effect was not powerful enough to result in mortality differences. Conventional outcomes were selected to comply with standard definitions and to be able to compare our results with other studies. Objective outcomes provided a more realistic analysis of the effect of prophylactic antibiotics on different variables.
Despite the lack of randomization, the complete lack of significant differences or even remarkable clinical trends in 21 of 22 outcome variables makes the likelihood of a β statistical error rather small. Our results confirm the widespread belief that the only effect of more antibiotics is a higher incidence of bacterial resistance.23 The presence of resistant infection was associated with longer SICU and hospital stays and with a higher incidence of death. Prolonged antibiotic prophylaxis by multiple agents was an independent risk factor for the development of bacterial resistance. Although it is difficult to establish that prophylactic antibiotics, resistant infections, and longer hospital stays or death are linked, such a cause-effect relationship is intuitively possible. Our subanalyses, aimed at identifying groups of patients who may benefit from receiving more antibiotics, failed to find such a group but were limited by small sample sizes.
Although other investigators have shown a decrease in wound infections due to an improved enterococcal coverage provided by a combination of ampicillin and sulbactam compared with cefoxitin,24 we could not confirm this finding because of the low prevalence of enterococcal infections in our SICU and the inadequate sample size.
Finally, it is not clear whether the two methods of antibiotic prophylaxis were equally effective or ineffective. Prolonged administration does not compensate for inadequate dosing or inappropriate selection of antibiotics. The standard dosing of frequently used antibiotics (eg, cefoxitin sodium, 2 g every 6 hours, or gentamicin sulfate, 1 mg/kg every 12 hours) may be inadequate in patients who have bled extensively, received massive resuscitation, and been subjected to bacterial insults not adequately covered by the specific agents.25,26 It is possible that rather than abolishing the idea of more antibiotics, we might consider aggressive prophylaxis with high doses of powerful antibiotics in selected groups of trauma patients. Evidence exists of the increased effectiveness of using higher rather than standard doses to prevent infection in trauma patients, but such evidence is still inadequate.10,25,26
We conclude that in the examined population, prophylaxis administered longer than 24 hours with multiple antibiotics does not improve morbidity and mortality or shorten hospital stay compared with prophylaxis administered for 24 hours with 1 broad-spectrum antibiotic, and may increase bacterial resistance. Severe trauma should not be a reason for more prophylactic antibiotics at the standard doses or in combinations, as are used currently.
This study was presented at the 109th Scientific Session of the Western Surgical Association, San Antonio, Tex, November 12, 2001.
Corresponding author: George C. Velmahos, MD, PhD, LAC/USC Medical Center, 1200 N State St, Room 9900, Los Angeles, CA 90033 (e-mail: email@example.com).
John A. Weigelt, MD, Milwaukee, Wis: First, I want to thank George for giving me the paper in plenty of time before the meeting and also for a great presentation. His message is simple. More is not better. It is not a new message, but one we should listen to again and maybe even heed this time. I like his message, but I have a number of questions regarding his scientific methods and why he arrived at his conclusions.
First, why should prophylactic or presumptive antibiotics have any effect on your outcome measures? To my knowledge, these agents are given to prevent surgical wound infection and do no more. Surgical wound infection is not even one of your primary or secondary end points in this study. If you are asking us to change our paradigm regarding perioperative antibiotics, do you have any data supporting any of their effects on any of your outcome measures?
Second, resistant organisms increased with multiple antibiotics. Another good point that we need to remember but is also not new. However, you never defined for us resistant organisms in your presentation or in your manuscript. What were they? Were they more common with some antibiotic combinations, and, most importantly, tell us, please, what we can learn about preventing resistance from your data other than short antibiotic use.
Finally, one comment. I question your manuscript's discussion that targeted antimicrobial prophylaxis is neither helpful nor possible. All surgeons caring for injured patients must work as a team and decide what is best for the patients, not for themselves. Short courses of specific antibiotics for specific injuries are one of the best patient practices we need to implement.
Frederick A. Moore, MD, Houston, Tex: It was obvious in your group comparisons that the SHORT antibiotic group had less complex orthopedic patients. Did you include complex orthopedic patients in your logistic regression models? If you were to eliminate the long antibiotic descriptor from the models, would complex orthopedic injuries emerge as independent risk factors for adverse outcomes? I suspect that long antibiotic therapy simply describes patients with complex orthopedic problems.
Basil A. Pruitt, Jr, MD, San Antonio, Tex: The authors state that there was an increase in antibiotic resistance in the prolonged treatment group, but it would be helpful to know whether that resistance was to the antibiotics given or to antibiotics in general. Prolonged antibiotic prophylaxis was also said to be associated with a greater risk of death. To assess the importance of that association we need to know whether the patients died of infection and the cause of death of all patients who died. The authors also noted that the patients receiving prolonged antibiotic prophylaxis had longer ICU and longer hospital stays, but was that simply because they were sicker or more severely injured? Lastly, prolonged prophylaxis appears to have been employed predominantly in patients with orthopedic injury. If that is true and you delete the orthopedic patients from the comparisons, do the same outcome differences persist?
Ronald G. Latimer, MD, Santa Barbara, Calif: A question to the author about the paper's title. If antibiotics are given after injury, is it truly prophylaxis? Shouldn't we be discussing the therapeutic use of antibiotics rather than prophylaxis?
James E. Goodnight, Jr, MD, PhD, Sacramento, Calif: Dr Velmahos, your study begs the question, should you give "prophylactic antibiotics" at all?
Dr Velmahos: You asked if prophylactic antibiotics are meant to prevent only against wound infection? Well, there are many studies in trauma that use outcomes such as intra-abdominal infection or extra-abdominal infection and not just wound sepsis. By this study, we respond to the general feeling of many surgeons who believe that antibiotics are a panacea against all type of infections.
You asked about the definition of resistant infection. The way we defined it was by the growth of resistant organisms in more than 3 antibiotics and a need to treat a clinically sick patient. Although we have not analyzed in this study the bacteriology of our cultures, we did not observe an increase in enterococcal infections. Dr Weigelt has advocated that ampicillin-sulbactam would be a better prophylactic antibiotic than cefoxitin because it provides better coverage against Enterococcus. However, the prevalence of such infections in our SICU is not high.
With regard to targeted antibiotics and their role, I did not imply that targeted antibiotics have no role. The appropriate antibiotic or antibiotic combination at the appropriate doses may be useful. Unfortunately, the antibiotics that we are giving in the doses administered are most probably useless in these patients with the increased volumes of distributions, multiple blood transfusions, whole blood volume washouts, and profuse immunocompromise.
Dr Moore and Dr Pruitt asked about the results of our analysis if we excluded orthopedic patients. When we did this, the groups still remained similar with regard to infectious morbidity and organ failures. However, the orthopedic patients carried most of the weight of the difference in resistant infections. It was these patients who were mostly responsible for antibiotic resistance. It seems that our orthopedic colleagues treat fractures with inadequate antibiotics leading only to resistance.
Dr Pruitt asked whether resistance was generic or specific to particular antibiotics. Again, we have not fully analyzed the bacteriology of our cultures, but I can tell you that resistant bacteria grew essentially against all antibiotics that were used prophylactically.
Dr Latimer, you argue against the term prophylactic and propose it should be rather therapeutic. I agree that by the strict definition of prophylaxis, antibiotics can never be prophylactic in trauma. Other authors have used the term presumptive. Due to the lack of an appropriate definition, we used the current one to comply with the rest of the literature.