Background: The purpose of the present study was to perform a systematic review and meta-analysis of the use of alternative antibiotic regimens—including (A) antibiotic prophylaxis versus no prophylaxis, (B) longer versus shorter duration of antibiotic prophylaxis, and (C) alternative drugs—for patients with open fracture of the extremities.
Methods: Data sources included CINAHL, EMBASE, MEDLINE, the Cochrane Central Registry of Controlled Trials (CENTRAL), and the Cochrane database of systematic reviews from 1965 to December 2013. All randomized controlled trials comparing the effectiveness of antibiotic prophylaxis in patients with open fracture of the extremities were eligible.
Results: We identified 329 potentially eligible articles, of which seventeen proved to be eligible. In four randomized controlled trials involving 472 patients, we found a significantly lower infection rate in patients receiving antibiotic prophylaxis compared with those not receiving antibiotic prophylaxis (risk ratio = 0.37 [95% confidence interval, 0.21 to 0.66]; absolute risk reduction = 9.6% [95% confidence interval, 5.2% to 12.1%]). In three studies involving 1104 patients, we found no difference in the infection rate when a longer duration of antibiotics (three to five days) was compared with a shorter duration (one day) (risk ratio = 0.97; 95% confidence interval, 0.69 to 1.37). Confidence in the estimates for both questions was low to moderate. Individual comparisons of alternative drugs yielded estimates warranting only low to very low confidence.
Conclusions: Results of randomized controlled trials performed to date provide evidence that antibiotic prophylaxis reduces subsequent infection and that courses as short as one day are as effective as courses of three to five days, although the evidence warrants only low to moderate confidence. Given current practice, a large, multicenter, low risk of bias, randomized controlled trial enrolling representative populations and addressing the duration of antibiotics may be the next optimum step in investigation.
Level of Evidence: Therapeutic Level I. See Instructions for Authors for a complete description of levels of evidence.
Open fractures—fractures involving the communication of bone through the skin and with the external environment1—are considered to be contaminated wounds. A key issue in treatment is the prevention of infection, which can occur in association with as many as 50% of open fractures. Strategies to minimize infection include copious irrigation, debridement, optimum fracture repair, and administration of antibiotics2-4.
Despite the implementation of these strategies, bacterial infection remains an important cause of nonunion and osseous instability after open fracture2. According to the results of culture of tissues at open fracture sites prior to orthopaedic operations, gram-positive bacteria are most common in association with Gustilo-Anderson type-I fractures5,6. The more severe the fracture type (i.e., Gustilo-Anderson type II or III), the more likely that gram-negative or mixed bacteria are present7.
Given the certainty of contamination and the very high incidence of infection, prophylactic use of antibiotics after open fracture has become routine8,9. Depending on the severity of the fracture and the preferences of the treating surgeon, the antibiotic may be given orally, parenterally, or locally. Surgeons frequently use multiple antibiotics to decrease the risk of resistance and to increase efficacy10. Some authorities have recommended that patients with open fractures of the extremities should receive intravenous antibiotics as prophylaxis against infection as early as possible, preferably within three hours after the injury9,11-13. Commonly recommended regimens include first or second-generation cephalosporins or, if the patient has an allergy to such agents, clindamycin11-14.
However, debate continues with regard to the duration of use, the route of administration, and the choice of antibiotics. Therefore, we undertook a systematic review and meta-analysis of the available evidence from randomized controlled trials to assess (A) the impact of antibiotic prophylaxis versus no prophylaxis on the rate of infection, (B) the impact of longer versus shorter duration of antibiotic prophylaxis on the rate of infection, and (C) the impact of alternative antibiotics on the rate of infection.
Materials and Methods
We included randomized clinical trials in which (1) the patients had presented with one or more open fractures involving the arms and legs, (2) an antibiotic prophylactic regimen was compared with any other regimen or with no antibiotic prophylaxis, and (3) the incidence of postoperative infection was reported as one of the clinical outcomes. We excluded studies addressing the use of antibiotics in patients with known infections; studies of patients with fractures involving the fingers or toes; and studies restricted to patients with gunshot wounds, injuries from bomb explosions, or HIV/AIDS (human immunodeficiency virus/acquired immune deficiency syndrome).
Data Sources and Search Strategy
We searched CINAHL, EMBASE, MEDLINE, the Cochrane Central Registry of Controlled Trials (CENTRAL), and the Cochrane database of systematic reviews from 1965 to December 2013. We restricted the search to human participants. Keywords included antibiotics, antimicrobial, antibiotic prophylaxis, open fracture, compound fracture, Gustilo-Anderson type, fracture fixation, nonunion, infection, and the names of specific antibiotics (see Appendix). An expert librarian (N.B.) developed the search strategy for our systematic review. One can retrieve the same search results in our study by entering in the MEDLINE search engine all of the commands listed in the Appendix and by using the translated approach with the same keywords and logic from the Appendix in other literature databases listed above16.
Study Selection and Data Abstraction
Reviewers, working in pairs, independently screened titles and available abstracts of identified citations and adjudicated the eligibility of the full text of titles and abstracts that were judged to be potentially eligible.
Using a data abstraction form created with Microsoft Excel, teams of reviewers, working in pairs, extracted the following data independently from each eligible study: funding and country of study; duration of follow-up; population characteristics, including sex distribution and age; sites and Gustilo-Anderson types of open fractures5,6 in terms of both numbers and proportions; intervention data, including antibiotic prophylaxis intervention details (antibiotic drug, dose, route of administration, start time, duration) and fluid irrigation type; and outcome data, including the number and proportion of infections in each study group and the risk ratio (RR) and its 95% confidence interval (CI).
Assessment of Risk of Bias and Confidence in Effect
Using a modified version of the Cochrane risk of bias tool17,18, teams of reviewers, working in pairs, independently assessed seven domains: (1) adequacy of sequence generation; (2) allocation concealment; (3) blinding of participants, health-care professionals, data collectors, and data analysts; (4) blinding of outcome assessors; (5) incomplete outcome data; (6) selective outcome reporting; and (7) other sources of bias. Reviewers chose from response options of “definitely yes,” “probably yes,” “probably no,” and “definitely no” for each of the domains, with “definitely yes” and “probably yes” ultimately assigned a low risk of bias and “definitely no” and “probably no” assigned a high risk of bias17. We used the “risk of bias summary” figure function of Review Manager (RevMan [Computer program]. Version 5.2. Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, 2012) to present the risk of bias result of individual studies from the seven domains of the modified Cochrane risk of bias instrument17,19. The RevMan software generates a red dot with a minus mark when we enter “high risk of bias” for a domain in a particular study and a green dot with a plus mark when we enter “low risk of bias.”19 The resulting figure provides a vivid and transparent accounting of the risk of bias for every included study (Fig. 2)19.
Reviewers resolved disagreements regarding eligibility, data abstraction, or risk of bias through discussion18.
The GRADE (Grading of Recommendations Assessment, Development and Evaluation) methodology was used to rate confidence in estimates of effect (quality of evidence) as high, moderate, low, or very low18,20. We used detailed GRADE guidance to assess overall risk of bias18,21, imprecision18,22, inconsistency18,23, indirectness18,24, and publication bias18,25 and summarized results in an evidence profile18,26. In terms of the risk of bias across studies, there are no serious limitations when most information comes from studies at low risk of bias and therefore we do not rate down for risk of bias. There are serious limitations when most information comes from studies with a moderate or high risk of bias, and under these circumstances we rate down one level in the extent of risk of bias21. Imprecision relates to the 95% CI around the effect difference between the two groups being compared, with consideration of effect size and sample size22. The quality of a body of evidence may vary from high (four plus; ⊕⊕⊕⊕) to moderate (three plus; ⊕⊕⊕○) to low (two plus; ⊕⊕○○) to very low (one plus; ⊕○○○), depending on the overall rating of the studies included in the meta-analysis27. We included as footnotes detailed explanations of our judgments in the GRADE evidence profile26.
Data Synthesis and Statistical Analysis
We assessed chance-corrected agreement in full-text eligibility judgments and risk of bias assessments using the kappa statistic28. We calculated pooled risk ratios and associated 95% confidence intervals for infection with use of random effects models applying the Mantel-Haenszel (M-H) method29. We calculated the risk difference and 95% confidence intervals on the basis of the values of the baseline risk and the risk ratio18,19. Specifically, we used the median infection rate of the control groups as the baseline risks. In the case of four studies, the median was the average of the rates in the middle two studies. The heterogeneity among the studies was assessed with the I2 statistic19. Analyses were performed using RevMan version 5.2.
Our search identified 13,499 abstracts from the electronic database search, of which 1155 were excluded as duplicates and an additional 11,972 were excluded on the basis of review of the title and abstract (Fig. 1). We were unable to access the full text of forty-three articles that had been deemed potentially eligible on the basis of a review of the title and abstract. Of the 329 articles for which the full text was reviewed, 276 were not randomized controlled trials and thirty-six did not assess open fractures involving the extremities, leaving seventeen eligible randomized controlled trials (Fig. 1)30-46.
The agreement regarding full-text eligibility selection was excellent (kappa, 0.72), and the agreement in risk of bias assessments was near perfect (kappa, 0.85 to 1.0).
These studies allowed us to compare (A) antibiotic prophylaxis versus no prophylaxis, (B) longer versus shorter duration of antibiotic prophylaxis, and (C) miscellaneous antibiotic regimens (Table I).
Antibiotic Prophylaxis Versus No Prophylaxis
Four randomized controlled trials enrolling 472 patients addressed fracture site infection with and without antibiotic prophylaxis30-33. Interventions included penicillin and first-generation cephalosporins administered intravenously. In two studies, the patients in the control groups were given no antibiotic prophylaxis and the patients in the intervention groups received antibiotics parenterally for five days30,32. In two studies, the patients in the control group received placebo31,33 and the patients in the intervention group received intravenous antibiotics for two days31 or intravenous antibiotics for four days followed by oral antibiotics for six days33. We found that in all four trials, randomization sequence generation and blinding of outcome assessment had high risk of bias17,30-33. Three of the four trials had high risk of bias on the basis of allocation concealment17,30,32,33. Risk of bias was also high because of lack of blinding of participants and health-care providers in two of the four trials17,30,32 (Fig. 2).
Results suggested a large, consistent reduction in infection risk with antibiotic use (RR = 0.37 [95% CI, 0.21 to 0.66], I2 = 0%) (Fig. 3), resulting in a risk difference of 9.6% fewer (95% CI, 5.2% to 12.1% fewer) infections in the antibiotic prophylaxis groups than the comparator groups. We rated the confidence in the estimates as low to moderate because of risk of bias and imprecision (Table II)21-27. We rated down imprecision not because of the width of the confidence interval but rather because of the small number of events, resulting in a failure to meet optimum information size criteria22.
Longer Versus Shorter Duration of Antibiotic Prophylaxis
Three randomized controlled trials enrolling 1104 patients were included in the meta-analysis comparing the rates of infection after longer-duration (three to five days) and shorter-duration (one day) prophylactic antibiotic use34-36. Interventions included first or second-generation cephalosporin given continuously for three to five days34-36. Controls included single-dose cephalosporin34, double-dose cephalosporin34, and one day35 of second-generation cephalosporin or single-dose fluoroquinolone36. In all three studies, we judged allocation concealment and the lack of blinding of outcome assessment as having a high risk of bias17,34-36. Randomization sequence generation (two of three trials)17,34,35 and selective reporting (one of three trials)17,35 were also judged as having a high risk of bias. The risk was low for attrition, reporting, and other bias17,34-36 (Fig. 2).
Results showed very similar rates of infection between groups receiving three to five days and one day of antibiotics (RR = 0.97 [95% CI, 0.69 to 1.37], I2 = 0%) (Fig. 4). We rated the quality of evidence as low to moderate because of the risk of bias and imprecision (Table II)21-27.
Individual Comparisons of Antibiotic Regimens
The risk of bias in individual studies was generally high, mainly because of inadequate sequence allocation, lack of allocation concealment, and lack of blinding (Fig. 2). The risk of bias was high in all studies17,37-44 except for those of Johnson et al.17,45 and Moehring et al.17,46, which had a moderate risk of bias. Studies compared a variety of different regimens; confidence intervals in most studies were very wide37-46 (Table I).
Two of the ten studies demonstrated significant differences between antibiotic treatments38,39. Vasenius et al. reported that patients in the clindamycin group had a lower infection rate than patients in the cloxacillin group (RR = 0.46 [95% CI, 0.24 to 0.91])38. Waikakul et al. showed that patients in the ofloxacin group had a higher infection rate than patients in the dicloxacillin group (RR = 0.19 [95% CI, 0.04 to 0.77])39. We rated confidence in the estimates in all studies as low or very low because of the risk of bias and imprecision21-25,27,37-46.
Results from four randomized controlled trials30-33 enrolling a total of 472 patients suggested large reductions in the relative risk (RR = 0.37 [95% CI, 0.21 to 0.66]) and absolute risk (9.6% [95% CI, 5% to 12%]) of infection with antibiotic prophylaxis versus no prophylaxis, although the results warrant only low to moderate confidence due to limitations of risk of bias (Fig. 2) and small total sample size and number of events (Fig. 3, Table II). A meta-analysis of three studies34-36 enrolling 1104 patients suggested no difference in the infection rate when a longer duration of antibiotics (three to five days) was compared with a shorter duration (one day) (RR = 0.97 [95% CI, 0.69 to 1.37]), although confidence again was low to moderate because of the risk of bias (Fig. 2) and wide confidence intervals (Fig. 4, Table II). Individual comparisons37-46 all were associated with low or very low confidence in the estimates of effect because of the risk of bias (Fig. 2) and imprecision.
The strengths of the present study include explicit eligibility criteria, a comprehensive search for relevant randomized controlled trials in all languages, duplicate assessments of eligibility and risk of bias with a high level of agreement, and application of GRADE criteria for confidence in estimates of effect. The weaknesses of the present study are related to limitations in the evidence, including the small number of eligible studies, the relatively small sample size, and the high risk of bias in most studies. These limitations led to our relatively low ratings of confidence in the estimates. Among the studies34-36 pooled in Figure 4, the unit of analysis differs, with two of the studies35,36 comparing infection rates of longer versus shorter durations of antibiotic prophylaxis on the basis of the number of patients and with the other study using the number of fractures34. We contacted the first author of the latter study to obtain the number of infection events using patients as the unit of analysis. However, the original data for that study were not retrievable. We did not exclude that study in the pooled result because, with the small number of events, the difference of using either fractures or patients as the unit would not have altered our result.
Our finding regarding the infection-preventing effect of using antibiotic prophylaxis for patients with open fractures is consistent with the findings of previous systematic reviews14,47-49. Gosselin et al., in an analysis of eight studies involving 1106 participants, found that antibiotics protected patients from early infection when compared with no antibiotics or placebo (RR = 0.43 [95% CI, 0.29 to 0.65])14. Our study included fewer randomized controlled trials than did the study by Gosselin et al. because we excluded studies involving fractures of the fingers and gunshot wounds. We excluded studies of finger fractures because the prognosis following such fractures differs from that following extremity fractures. For finger fractures, the anatomy of the hand is such that, if an infection occurs, the infection may track down the tendon sheath50. For gunshot wounds, optimum management may vary depending on whether the gun was a high-velocity weapon, a low-velocity weapon, or a shotgun, each of which carries its own prognosis51.
Other systematic reviews have led to the same conclusion that antibiotic prophylaxis has protective effects but have not provided the pooled risk ratios for the included studies47-49. To our knowledge, no previous systematic review has compared the effects of different durations of antibiotic prophylaxis or different antibiotic regimens. Also, to our knowledge, no previous review has used synthetic approaches such as GRADE to assess the quality of evidence27,52.
The low confidence in the effect of prophylaxis, despite the apparently large effect, suggests that the conduct of additional large studies with designs ensuring low risk of bias would be desirable. The long-established consensus in the community in favor of antibiotics would likely make such a trial unfeasible in the current environment. Our results suggest, however, that a single day’s exposure to antibiotics is similar in its effects to more prolonged administration. An initial trial comparing shorter and longer regimens might well be feasible, and, if it showed no difference, the climate of opinion regarding the possibility of a trial of antibiotics versus no antibiotics could possibly change.
Aside from the issues of whether antibiotics should be administered at all and the duration of their use, other issues, including the optimum choice of antibiotic, the route of administration, and the optimum start time, remain unresolved. It is also possible that different regimens and durations may be preferable for different patients according to the site and severity of the fractures. Previous in vitro and in vivo experimental studies in both animals and humans showed negative effects of topical antibiotics (in the form of impregnated beads or cements in clinical practice) on bone cell function and fracture repair. High concentrations of most antibiotics can inhibit bone-healing by affecting chondrocytes and osteoblasts and increasing mineralization of bone. The optimum dose of topical antibiotics to prevent infection without negative effects on bone repair is unknown53.
Resolution of these issues will require randomized trials of superior design to those undertaken to date. Such studies should address the limitations of previous studies, including the failure to document adverse reactions to antibiotics and the nature of the infections that occur (e.g., superficial or deep). In addition, they should be large, multicenter studies that are sufficiently powered to provide definitive results. Finally, they should implement strategies to reduce the risk of bias (including concealed allocation and blinding of patients, clinicians, and those involved in outcome assessment) as well as strategies to minimize loss to follow-up.
The authors thank the following colleagues for their contribution to this research: Mohamad Alshurafa, Lisa Cronin, Nicole Simunovic, Yanping Zhao, Bill Ristevski, Arnav Agarwal, Michal Seweryn, Kamran Naseem, Thomas Agoritsas, Arnaud Merglen, Ingmarie Skoglund, Aurelia Desplain, Andy Radoslav, Hejia Song, Mark Loeb, Paul Elias Alexander, Nigar Sekercioglu, Li Wang, Qi Zhou, Toshiaki A. Furukawa, Souzan Mirza, Regina Kunz, and Victor Wang.
Appendix 1: MEDLINE Title and Abstract Search Strategy for Effects of Antibiotic Prophylaxis in Patients with Open Fracture
1. antibiotics.mp. or exp Anti-Bacterial Agents/
2. antibiotic prophylaxis.mp. or exp Antibiotic Prophylaxis/
3. (anti-microb* or anti bact* or antibact*).mp.
4. (antibiotic* or antimicrob*).mp.
8. exp AMOXICILLIN/ or AMOXICILLIN.mp.
9. exp Ampicillin/ or AMPICILLIN.mp.
10. CLAVULANIC ACID.mp.
14. exp Cephalosporins/ or CEPHALOSPORIN*.mp.
15. (KEFLEX or CEFAMANDOLE or KEFADOL or CEFAZOLIN* or KEFZOL or CEFIXIME or SUPRAX).mp.
16. (CEFOTAXIME or CLAFORAN or CEFOXITIN or MEFOXIN or CEFPIROME or CEFROM or CEFPODOXIME).mp.
17. (ORELOX or CEFPROZIL or CEFZIL or CEFRADINE or VELOSEL or CEFTAZIDIM or ORTUM or KEFADIM).mp.
18. (CEFTRIAXONE or ROCEPHIN or CEFUROXIME or ZINACEF or ZINNAT or CEFONICID or AZTREONAM).mp.
19. (AZACTAM or IMIPENEM or ILASTATIN or PRIMAXIN or MEROPENEM).mp.
20. (TETRACYCLINE* or DETECLO or DEMECLEOCYCLIN or LEDERMYCIN or DOXYCYCLINE or VIBRAMYCIN).mp.
21. (MINOCYCLINE or MINOCINE or OXYTETRACYCLINE or TERRAMYCIN or MACROLIDE*).mp.
22. (ERYTHROMYCIN or ERYMAX or ERYTHROCIN or ERYTHROPED or AZITHROMYCIN or ZITHROMAX).mp.
23. (CLARITHROMYCIN or KLARICID or TELITHROMYCIN or KETEK or TRIMOXAZOLE or SEPTRIN).mp.
24. (TRIMETHOPRIM or MONOTRIM or TRIMOPAN or METRONIDAZOLE or FLAGYL or METROLYL).mp.
25. (PHENOXYMETHYLPENICILLIN or SULFAMETHOXAZOLE or OXACILLIN or CEPHALOTHIN or SULBACTAM).mp.
26. (OFLOXACIN or CLINDAMYCIN or GENTAMYCIN or VANCOMYCIN).mp.
27. (CEFACLOR or DISTACLOR or CEFADROXIL or BAXAN or CEFALEXIN or CEPOREX).mp.
28. (TIMENTIN or FLUCLOXACILLIN or FLUAMPICIL or MAGNAPEN or PIPERACILLIN or TAZOCIN).mp.
29. (streptomycin or cefalotin or dicloxacillin).mp.
31. exp fractures, open/
32. (orthopedic adj2 surg*).mp.
33. (open adj9 fracture*).mp.
34. ((open adj2 reduction) and fracture*).mp.
35. (Gustilo or Gustillo).mp.
36. anderson type*.mp.
37. (compound adj9 fracture*).mp.
38. ununited fractures.mp. or exp Fractures, Ununited/
39. fracture fixation.mp. or exp Fracture Fixation/
40. fracture*.mp. or exp Fractures, Bone/
41. (infect$ adj3 (bone$ or fracture$)).mp.
42. ((nonunion or non union) adj9 fracture*).tw.
44. 30 and 43
45. animals/ not humans/
46. 44 not 45
Source of Funding: No external funds were received in support of this study.
Investigation performed at the Department of Clinical Epidemiology and Biostatistics, McMaster University Health Sciences Centre, McMaster University, Hamilton, Ontario, Canada
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. None of the authors, or their institution(s), have had any financial relationship, in the thirty-six months prior to submission of this work, with any entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, no author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.
- Copyright © 2015 by The Journal of Bone and Joint Surgery, Incorporated