Introduction

Objective

The purpose of this study is to establish whether the isolates associated with clinical Gram negative burn wound infection differ between burn centres. This study also sought to establish standardised incidence rates for the organisms identified.

Methods

Results

4.1 Literature Retrieval and Critical Appraisal

Thirty four studies were retrieved by the initial literature search (Figure 1). Narrative critical appraisal is provided in extensive detail in File S7. Of these studies, 20 did not conform to the inclusion criteria and were excluded (File S6). Fourteen studies were included for critical appraisal (Table 1). A further 5 studies did not conform to operational definitions (section 3.3). “Strength of the evidence” underpinning the remaining studies was evaluated by critical appraisal and is reported in Table 2. Two further studies [22], [23] investigated primary incidence over week 1 post-burn only. Because of this methodological heterogeneity, their results could not be pooled or confidently compared to the rest of the literature. Standardised incidence data from the remaining 7 studies (Table 2) was extracted and pooled (Figure 2). The geographic location of the burn centres from which the data was pooled is reported in Figure 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. PRISMA-style scheme reporting the literature retrieval and selection strategy arriving to the final 7 studies whose data could be pooled for statistical analysis.

https://doi.org/10.1371/journal.pone.0095042.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Box-whisker plot reporting data dispersion for the standardised incidence of Gram-negative burn wound injury in civilian adult hospitalised patients.

Data shown represents mean±1SD.

https://doi.org/10.1371/journal.pone.0095042.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Geographic distribution of the centres from which studies were critically appraised.

The study design and sample size are provided in figure 2.

https://doi.org/10.1371/journal.pone.0095042.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Literature Included for Critical Appraisal.

https://doi.org/10.1371/journal.pone.0095042.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Incidence of Gram negative organisms causing clinical burn wound infection extracted from eligible primary literature.

https://doi.org/10.1371/journal.pone.0095042.t002

Discussion

This study studied primary literature to establish whether the isolates associated with clinical Gram negative burn wound infection differ between burn centres. Burn wound infection has traditionally been a difficult area for the clinician, and one of the primary reasons is the difficulty in knowing which organisms to target. This study established that organisms causing Gram-negative burn wound infection do not differ significantly between burn centres. Analysis of the standardised data confirm, within this study’s limitations, the principal hypothesis that the organisms causing Gram-negative burn wound infection are similar, regardless of geographic location of the treating centre (p<0.05). This finding presents a significant departure from traditional teaching regarding the behaviour of infected burn wounds (section 1). The lack of standardised data reporting may be one possible explanation for this discord [25]. In contrast, aetiological profiles and incidences of the major protagonists in other disease such as urinary tract infection have long been established, and are constantly reviewed [6], [7]. Establishing the organisms that commonly cause Gram-negative infection in burn wounds may confer to acute burns patients the long-term benefits enjoyed by these other common diseases, such as rationalised, expedited, targeted antimicrobial development, prescribing, infection control, and surveillance of resistance patterns. These findings, integrated with local data regarding susceptibility patters, may facilitate the formulation of “first line” antibiotic treatment strategies, and increase probabilities of therapeutic efficacy. For example, worldwide resistance to commonly used broad spectrum antibiotics such as ceftriaxone is common (from 16.3% E. coli to 64% of Acinetobacter strains studied) [26].

Identifying this set of bacteria may also have implications on defining research strategies in drug redevelopment. Only up to 1.4% of isolates from across the species identified are resistant to colistin, emphasising the usefulness of this antibiotic as a drug of last resort [26]. These observations lend further credence to the strong interest of this venerable class’s redevelopment via semi-synthetic chemistry approaches. The approach to identifying a common set of bacteria responsible for burn wound infection has already borne translational fruit. Recently, an in-depth analysis for commonalities of biochemical and virulence mechanisms involved in the aetiology of infection with these organisms identified that substantial production of bradykinin is common to all these pathogens and leads to enhanced vascular permeability and sequestration of macromolecules. Based on this principle, we recently published evidence of a novel, size-based paradigm for drug targeting in infection [27]; statistically optimised a bioresponsive polymeric payload carrier to achieve this goal [28]; proposed a novel class of macromolecular antimicrobial agents capable of locally triggered enzymatic activation at the infected site, retaining antimicrobial potency to match the conventional clinical equivalent whilst significantly reducing in vivo toxicity [29], [30].

Ample evidence exists to support the notion that morbidity, mortality and quality of life outcomes in burn patients is associated to the organisms identified in this study. For example, wound infection with P. aeruginosa, E. coli, and K. pneumoniae wound is an independent predictor of mortality [3], [31]. These bacteria also promote failure of healing [32] which is of major consequence to the management of extensive burn wounds by serial excision where time to healing is essential and donor sites for grafting come at a premium. Moreover, specific risk factors associated to burn wound infection with these organisms have been identified [33]–[35], and modification of management practices may lower infection rates resulting in improved outcomes. The importance of identifying this set of organisms as the prime perpetrators of Gram-negative burn wound infection, regardless of the treating centre, is therefore apparent.

P. aeruginosa (Figure 2) exhibits an interesting, wider dispersion of dispersion of data compared to the other organisms. A possible explanation lies in the presence of an outlier data set [11] (Table 2), hence the long superior whisker for the P. aeruginosa plot in Figure 2. However ANOVA, is remarkably robust to moderate departures from normality caused by outlier data sets. One possible explanation for this outlier may lie in the susceptibility of the standardised incidence of P. aeruginosa to the overall infection rate reported from the relative burn centre. In fact, the overall infection rate reported from this centre is also high [11].

Such a study presented unique difficulties. Mere presence of organisms on a wound does not imply infection. Histological documentation of infection into viable tissue may secure the diagnosis. However, in practice, few if any centres worldwide have the substantial resources required to fulfil laboratory diagnostic criteria such as routine microbiological tissue histology and electron microscopy on a daily basis. Moreover, definitions of burn wound infection underpinned by these (tissue biopsy and electron microscopy) laboratory investigations are largely considered dated [36], [37]. As a working definition, an organism was defined as causative if it could be discerned from the primary literature that the organism/s was isolated from the wound in the presence of clinical infection. Clinical diagnosis of infection was based on reconciliation of primary literature data to the definitions of Greenhalgh et al. [21]. Studies including re-infection or relapse in their incidence rates were excluded. Whilst the clinical surrogates presented in these definitions may be less specific then exhaustive (but rarely performed) laboratory investigations, the approach presented herein would reduce the potential bias presented by the resources available to the researchers producing the primary literature.

As no RCTs were identified, the conduct of this review was adapted to consider the pooling of data from primary studies with heterogeneous methodologies. Systematic review of studies other than RCTs is well-established in the literature [21]. Therefore, a rigorous critical appraisal process qualified the “strength of the evidence” underpinning the primary data included in the statistical analysis. The critical appraisal process also facilitated the reconciliation of the primary literature to the operational definitions, ensuring comparability of the studies. Such an evidence-based approach is not new [12]. Quality of the evidence provided by the primary literature was limited (Table 2), justifying the use of an evidence-based methodology combined to the systematic review process. Since the burn centres in the primary literature were geographically separate and did not have overlapping catchment areas, it was reasonable to assume that “no interaction” occurred between the two independent variables studied.

The selection of adequate inclusion and exclusion criteria was challenging. A rigorous approach in data retrieval was observed and catalogued to prevent retrieval bias. In order to minimise the possibility data contamination with lurking variables, stringent criteria were applied. Evidence exists that delayed transfer may influence the bacteriological flora burned patients [38]. Studies concur that delayed presentation may alter the pathological flora on a burn wound [3], [39]. Military patients may pass through multiple facilities to the definitive site of treatment, and this delay presented a plausible lurking variable [38], [40]. This data was therefore excluded purely as a necessity to safeguard methodological rigour rather than implying a difference between military and civilian populations. In fact, data from Operation Iraqi Freedom and Operation Enduring Freedom suggests that similar bacterial flora is present in such patients, [38], [40], [41]. Indeed primary studies describing civilian injuries [22], [42] were also excluded for similar reasons.

A five to ten year cut-off for the retrieval of primary data is well-established in evidence-based methodology [12], [43]. Such time-limits for the inclusion of primary data in evidence-based medicine harks back to the main works of Archie Cochrane and DL Sackett [44], [45], affirmed in the Sicily statement for evidence-based research [17] and cited in multiple works since then [46]–[48]. Moreover until the 1980’s burn wounds were treated by the exposure method, with application of topical antimicrobials to the burn wound surface and gradual debridement with immersion hydrotherapy [24]. Thereafter, early burn wound excision and wound closure became the focal point of burn wound management, accompanied by a change from immersion hydrotherapy to showering hydrotherapy, and a consequent decrease in the rate of burn wound infection [24]. It is also well-established that early excision has reduced the incidence of invasive infection, as underscored by key publications immediately preceding our cut-off point for retrieval of primary literature. Pruitt et al. in particular, assert that the change to early excision and grafting significantly changed, both qualitatively and quantitatively, the incidence and identity of organisms causing these infections and their timing [38], [49].

It was assumed that by the year 2000, this method of treatment would have been well-established, coinciding with the cut-off point determined by the evidence-based methodology. However, it is acknowledged that such arbitrary cut-off points may have led to retrieval bias in selection of primary literature.

Conclusion

This study is the first to report that organisms causing clinical Gram-negative burn wound infection do not differ significantly between burn centres. Therefore, these findings establish a “target set” of Gram-negative pathogens for antimicrobial development and timely, effective treatment. P. aerugionsa, K. pneumoniae, E. coli, Enterobacter spp. and Proteus spp. were identified as the commonest Gram-negative pathogens to be isolated from clinically infected burn wounds regardless of the treating centre. It is of course acknowledged that other bacteria may infect the burn wound and it is especially important to monitor emerging infections [51]. However, we hope that finding a “target” set of pathogens may contribute to timely clinical treatment, effective, and clinically oriented antibiotic development. The threat posed by multi-drug resistant pathogens continues to increase, however research and clinical management in this specialist field remain poorly funded, fragmented and oftentimes reported in isolation. The paucity of relevant literature highlighted in this study illustrates the dire necessity for further epidemiological multi-disciplinary collaboration. We augur that the identification of a common aetiological Gram-negative burn wound profile will be a first step in this direction.

Supporting Information

Acknowledgments

The authors would like to thank Mr Stephen Honeyman, Mr Ian Malin, and the staff at the Welsh Burns Centre, Swansea, UK for their helpful comments. The authors would also like to thank the St. David’s Medical Foundation, Swansea University, for facilitating research into the conclusions of this study.