Antibiotics for preventing infection in open limb fractures



The use of antibiotics in the initial treatment of open fractures of the extremities is almost universal in all high income countries, where benefits are assumed to far out-weight potential risks. Since an open fracture is by definition contaminated, the use of antibiotics is therapeutic, not prophylactic, and aims at preventing subsequent infectious problems such as cellulitis, myositis, acute or chronic osteomyelitis, infected non-unions, recurrent abscesses, chronic drainage with fistula formation, and their associated impairment and disability.

In low income developing countries, antibiotics may not be used for various reasons: costs, lack of knowledge, low levels of suspicion or poor case-recognition from first-line health care workers, availability, accessibility, and the use of traditional or alternative health care. Even when antibiotics are available their use for open fractures may be delayed until the victim is seen in secondary or even tertiary care centres. This delay often far exceeds the accepted "golden period" of six to eight hours which is usually the standard of care in high income countries.

A recent systematic review (Gillespie 2002) concluded that antibiotic prophylaxis should be offered to those undergoing surgical treatment of closed fractures. To our knowledge no such review has been done for the use of antibiotics in open fractures. Our purpose is to review the evidence for use of antibiotics in the treatment of open fractures of the extremities.


To determine the evidence for the effectiveness of antibiotics in the initial treatment of open fractures of the extremities.

We will examine the following hypotheses:

1. Late infective and functional complications (cellulitis, myositis, acute or chronic osteomyelitis, infected non-unions, recurrent abscesses, chronic drainage with fistula formation, and their associated impairment and disability) are less frequent after administration within eight hours following open fracture.

2. Complications are similarly less frequent in those people to whom antibiotics are first administered more than eight hours following the fracture.

Criteria for considering studies for this review

Types of studies

Randomised and quasi-randomised (e.g. date of birth, alternation) controlled trials.

Types of participants

People of any age with open fractures of the extremities, including distal phalanxes, but excluding fractures of the skull, face and axial skeleton.

Types of intervention

Antibiotic administered before or at the time of primary treatment of the open fracture compared with placebo or no antibiotic. Trials comparing different antibiotics, different antibiotic dosages, route of administration or differences in timing or duration of administration will be excluded.

Types of outcome measures

  • early wound infection (as defined by each trial or, if not stated, 'draining wound')

  • chronic drainage

  • acute or chronic osteomyelitis

  • infected non-unions

  • amputations

  • deaths

Search strategy for identification of studies

See: Unavailable search strategy

We will search the Cochrane Musculoskeletal Injuries Group specialised register, the Cochrane Controlled Trials Register (The Cochrane Library current issue), MEDLINE (1966 to present) , EMBASE (1988 to present), LILACS (1992 to present) and reference lists of articles. Proceedings of meetings of the American Academy of Orthopedic Surgeons (AAOS), the Orthopedic Trauma Association (OTA) and the Societe Internationale de Chirurgie Orthopedique et Traumatologique (SICOT) will be reviewed. We will also contact published researchers in the field. In MEDLINE (OVID WEB) the following search strategy will be combined with all three stages of the optimal trial search strategy (Clarke 2001):

1. Antibiotic Prophylaxis/
2. exp Antibiotics/
3. (antibiotic$ or antimicrob$).tw.
4. or/1-3
5. Infection/
6. exp Wound Infection/
7. Sepsis/
8. infect$.tw.
9. or/5-8
10. Fractures, Open/
11. exp Fractures/
12. (open or compound).tw.
13. and/11-12
14. (infect$ adj3 (bone$ or fracture$)).tw
15. or/10,13,14
16. and/4,9,15

No language restriction will be applied.

Methods of the review

Two reviewers will independently screen each record for eligibility by initially examining titles, abstracts and key words. The full text identified by either reviewer for possible inclusion will be retrieved, and two reviewers will independently assess each for eligibility, assess methodological quality, and extract data from eligible reports using a standard proforma. Disagreements will be resolved by a third reviewer. We will extract data on the type of participants, the type of antibiotic used, the number randomised to intervention or control groups, the quality of allocation concealment, and the outcome measures stated in the protocol. We will write to the authors of reports if relevant information is missing. Trials in which we cannot confirm that random or quasi-random allocation had been used to allocate participants will be excluded.

Quality assessment

Methodological quality for each study will be independently assessed by at least two reviewers using a schedule derived from the generic evaluation tool developed by the Cochrane Musculoskeletal Injuries Group. Disagreements will be resolved by consensus or by the involvement of a third reviewer. The inter-rater agreement between reviewers will be noted. The scores are designed to give readers a general impression of trial quality and will not be used in any quantitative manner.

A. Was the assigned treatment adequately concealed prior to allocation?
1=states random, but no description, or quasi-randomisation (Category C)
2=small but real chance of disclosure of assignment (Category B)
3=method did not allow disclosure of assignment (Category A)

B. Were the outcomes of patients who withdrew described and included in the analysis (intention to treat)?
1=not mentioned
2=states numbers and reasons for withdrawal, but analysis unmodified
3=primary analysis based on all cases as randomised

C. Assessment of outcome. Were assessors of outcome blinded to treatment status?
1=not done or not mentioned
2=moderate chance of unblinding of assessors
3=action taken to blind assessors, or outcomes such that bias is unlikely

D. Were treatment and control groups comparable at entry?
1=large potential for confounding or not discussed
2=confounding small; mentioned but not adjusted for
3=unconfounded; good comparability of groups or confounding adjusted for

E. Was a placebo treatment assigned as part of the randomisation?

F. Were exclusion criteria clearly defined?
1=not defined
2=poorly defined
3=well defined

G. Was the method of assessment of wound infection stated?
1=not stated
2= clinical decision, or definite criteria without a microbiological diagnosis
3=definite criteria including a microbiological diagnosis

H. Was the surveillance for wound infection outlined?
1=not stated, or not active
2=active, but less than three months
3=active, and at least one year

For each intervention, we will estimate pooled relative risk in a fixed-effects model using Peto's modification of the Mantel-Haenszel method. We will calculate 95 per cent confidence intervals and two-sided p-values for each outcome. Selection bias will be assessed using Egger's weighted regression method and Begg's rank correlation test and funnel plot. Heterogeneity will be assessed using a Chi-square test at a five per cent significance level. Stratified analyses will be undertaken depending on whether antibiotics were given early (less than eight hours from injury) or late, and whether the trial was placebo controlled or not.

Potential conflict of interest

None known.


We thank the following for helpful comments at editorial review: Prof Marc Swiontkowski, Prof Rajan Madhok, Peter Herbison, Lesley Gillespie, Dr Janet Wale, and Mr John Stothard.

Sources of support

External sources of support

  • No sources of support supplied

Internal sources of support

  • No sources of support supplied