Dressings for the prevention of surgical site infection (Review)

Transcription

1 Dumville JC, Walter CJ, Sharp CA, Page T This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2011, Issue 7

2 T A B L E O F C O N T E N T S HEADER ABSTRACT PLAIN LANGUAGE SUMMARY BACKGROUND OBJECTIVES METHODS RESULTS Figure DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES CHARACTERISTICS OF STUDIES DATA AND ANALYSES Analysis 1.1. Comparison 1 Basic wound contact dressings compared with exposed wounds, Outcome 1 Proportion of wounds with SSI Analysis 2.1. Comparison 2 Advanced dressings compared with exposed wounds, Outcome 1 Proportion of wounds with SSI Analysis 3.1. Comparison 3 Comparisons between basic wound contact dressings, Outcome 1 Proportion of wounds with SSI Analysis 4.1. Comparison 4 Basic wound contact compared with film dressings, Outcome 1 Proportion of wounds with SSI Analysis 4.2. Comparison 4 Basic wound contact compared with film dressings, Outcome 2 Proportion of wounds with SSI: data pooled for clean wounds Analysis 4.3. Comparison 4 Basic wound contact compared with film dressings, Outcome 3 Pain associated with dressing (patient assessed) Analysis 4.4. Comparison 4 Basic wound contact compared with film dressings, Outcome 4 Patient acceptability.. 50 Analysis 5.1. Comparison 5 Basic wound contact dressings compared with hydrocolloid dressings, Outcome 1 Proportion of wounds with SSI Analysis 5.2. Comparison 5 Basic wound contact dressings compared with hydrocolloid dressings, Outcome 2 Proportion of wound with SSI - Holm 1998 raw data Analysis 5.3. Comparison 5 Basic wound contact dressings compared with hydrocolloid dressings, Outcome 3 No pain on removal Analysis 6.1. Comparison 6 Basic wound contact dressings compared with fibrous-hydrocolloid (hydrofibre) dressings, Outcome 1 Proportion of wounds with SSI Analysis 6.2. Comparison 6 Basic wound contact dressings compared with fibrous-hydrocolloid (hydrofibre) dressings, Outcome 2 Proportion of wounds with SSI - Vogt 2007 raw data Analysis 7.1. Comparison 7 Basic wound contact dressings compared with matrix hydrocolloid dressings, Outcome 1 No pain on removal Analysis 8.1. Comparison 8 Comparisons between advanced dressings, Outcome 1 Proportion of wounds with SSI.. 53 APPENDICES HISTORY CONTRIBUTIONS OF AUTHORS DECLARATIONS OF INTEREST SOURCES OF SUPPORT DIFFERENCES BETWEEN PROTOCOL AND REVIEW INDEX TERMS i

3 [Intervention Review] Dressings for the prevention of surgical site infection Jo C Dumville 1, Catherine J Walter 2, Catherine A Sharp 3, Tamara Page 4 1 Department of Health Sciences, University of York, York, UK. 2 Colorectal Surgery, Gloucestershire NHS Foundation Trust, Cheltenham, UK. 3 The Wound Centre, Sydney, Australia. 4 Royal Adelaide Hospital, Adelaide, Australia Contact address: Jo C Dumville, Department of Health Sciences, University of York, York, YO10 5DD, UK. jd34@york.ac.uk. Editorial group: Cochrane Wounds Group. Publication status and date: New, published in Issue 7, Review content assessed as up-to-date: 9 May Citation: Dumville JC, Walter CJ, Sharp CA, Page T. Dressings for the prevention of surgical site infection. Cochrane Database of Systematic Reviews 2011, Issue 7. Art. No.: CD DOI: / CD pub2. Background A B S T R A C T Surgical wounds (incisions) heal by primary intention when the wound edges are brought together and secured - often with sutures, staples, clips or glue. Wound dressings, usually applied after wound closure, provide physical support, protection from bacterial contamination and absorb exudate. Surgical site infection (SSI) is a common complication of surgical wounds that may delay healing. Objectives To evaluate the effects of wound dressings for preventing SSI in people with surgical wounds healing by primary intention. Search strategy We searched the Cochrane Wounds Group Specialised Register (searched 10 May 2011); The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2011 Issue 2); Ovid MEDLINE (1950 to April Week ); Ovid MEDLINE (In-Process & Other Non-Indexed Citations, May 9, 2011); Ovid EMBASE (1980 to 2011 Week 18); EBSCO CINAHL (1982 to 6 May 2011). There were no restrictions based on language or date of publication. Selection criteria Randomised controlled trials (RCTs) comparing alternative wound dressings or wound dressings with leaving wounds exposed for postoperative management of surgical wounds healing by primary intention. Data collection and analysis Two review authors performed study selection, risk of bias assessment and data extraction independently. Main results Sixteen RCTs were included (2578 participants). All trials were at unclear or high risk of bias. Nine trials included people with wounds resulting from surgical procedures with a contamination classification of clean, two trials included people with wounds resulting from surgical procedures with a clean/contaminated contamination classification and the remaining trials evaluated people with wounds resulting from various surgical procedures with different contamination classifications. Two trials compared wound dressings with leaving wounds exposed. The remaining 14 trials compared two alternative dressing types. No evidence was identified to suggest that any dressing significantly reduced the risk of developing an SSI compared with leaving wounds exposed or compared with alternative dressings in people who had surgical wounds healing by secondary intention. 1

4 Authors conclusions At present, there is no evidence to suggest that covering surgical wounds healing by primary intention with wound dressings reduces the risk of SSI or that any particular wound dressing is more effective than others in reducing the rates of SSI, improving scarring, pain control, patient acceptability or ease of dressing removal. Most trials in this review were small and of poor quality at high or unclear risk of bias. However, based on the current evidence, we conclude that decisions on wound dressing should be based on dressing costs and the symptom management properties offered by each dressing type e.g. exudate management. P L A I N L A N G U A G E S U M M A R Y No recommendations regarding type of wound dressing for the prevention of surgical site infection Millions of surgical procedures are conducted globally each year. The majority of procedures result in wounds in which the edges are brought together to heal using stitches, staples, clips or glue - this is called healing by primary intention. Afterwards, wounds are often covered with a dressing that acts as a barrier between them and the outside environment. One advantage of this may be to protect the wound from micro-organisms, and thus infection. Many different dressing types are available for use on surgical wounds, however, it is not clear whether one type of dressing is better than any other at preventing surgical site infection, or, indeed, whether it is better not to use a dressing at all. We conducted a review of all available, relevant, evidence regarding the impact of dressings on the prevention of surgical site infections in surgical wounds healing by primary intention. The review examined data from 16 randomised controlled trials and found no evidence to suggest either that one dressing type was better than any other, or that covering these wounds with dressings at all was better, at preventing surgical site infection, or that any dressing type improves scarring, pain control, patient acceptability or ease of removal. It is important to note that many trials in this review were small and of poor quality, at high or unclear risk of bias. Decisions on wound dressing should be based on dressing costs and the need for management of specific symptoms e.g., absorption of exudate. B A C K G R O U N D Description of the condition Millions of surgical procedures are conducted around the world each year. The majority of procedures result in surgical wounds that will heal by primary intention. This is where wound edges are re-approximated using sutures, staples, clips or glue, either alone, or in combination. Surgical site infection (SSI) is a common complication of surgical wounds that may delay healing. A US study found that in over 750,000 episodes of surgical hospitalisation, 1% resulted in an SSI (de Lissovoy 2009), and similar estimates have been found in France (Astagneau 2009). Furthermore, such values may underestimate the levels of SSI by not considering those that develop outside hospitals. In the UK it has been estimated that 4% to 5% of patients undergoing a surgical procedure contract an SSI (Health Protection Agency 2002; Smyth 2008). Whilst various patient factors can predict the likelihood of SSI, the type of surgical procedure performed exerts a major influence on risk. Surgical procedures involving clean body cavities have much lower infection rates, of around 3% to 5%, compared with procedures involving body cavities with infected, necrotic or dirty tissue - for example, colorectal surgery - which have surgical infection rates of around 10% to 30% (McLaws 2000). A widely used definition describing the contamination classification of surgical procedures is given below: Clean: Non-infective operative wounds in which no inflammation is encountered, and neither the respiratory, alimentary, genitourinary tract nor the oro-pharyngeal cavity is entered. In addition these cases are elective, primarily closed, and drained with closed drainage system when required. Clean/ Contaminated: Operative wounds in which respiratory, alimentary, genital or urinary tract is entered under controlled conditions and without unusual contamination. Specifically, operations involving the biliary tract, appendix, vagina and oropharynx are included in this category, provided no evidence of infection or a major break in sterile technique is encountered. Contaminated: Fresh, accidental wounds, operations with major breaks in sterile technique or gross spillage from the gastrointestinal tract, and incisions in which acute, non-purulent inflammation is encountered. Dirty: Old traumatic wounds with retained devitalised tissue and those that involve existing clinical infection or perforated viscera. 2

5 This definition suggests that organisms causing postoperative infection were present in the operative field before the operation. SSIs not only cause considerable patient morbidity, but also increase the consumption of health care resources. In the UK, the mean additional cost of treating an infected surgical wound (compared to a non-infected wound) was estimated at GBP 1618 (Plowman 2001), with much of this extra cost attributable to an increased length of hospital stay (mean increase of 6.5 days) (Plowman 2001). In the USA, de Lissovoy 2009 estimated that the extended length of stay and increased treatment costs associated with SSIs over a one-year period led to approximately 1 million additional inpatient-days, costing an additional USD 1.6 billion. Whilst SSIs can be difficult to define - one review identified 41 different definitions and 13 grading scales of SSI (Bruce 2001) - the Centers for Disease Control and Prevention (CDC) have published the following guidelines defining superficial and deep incisional SSIs (Horan 2008). A superficial SSI is defined as: an infection occurring within 30 days after the operation and only involving the skin and subcutaneous tissue of the incision that is associated with at least one of the following: Purulent drainage, with or without laboratory confirmation, from the surgical site. Organisms isolated from an aseptically-obtained culture of fluid or tissue from the surgical site. At least one of the following signs or symptoms of infection: pain or tenderness, localised swelling, redness or heat, and superficial incision is deliberately opened by the surgeon and is culture-positive or not cultured. A culture-negative finding does not meet this criterion. Diagnosis of SSI by the surgeon or attending physician. A deep incisional SSI is defined as: infection that occurs within 30 days after the operative procedure if no implant is left in place, or within one year if an implant is left in place, and the infection appears to be related to the operative procedure and involves deep soft tissues (e.g. fascial and muscle layers) of the incision associated with one of the following: Purulent drainage from the deep incision, but not from the organ/space component of the surgical site. A deep incision spontaneously dehisces (opens up) or is deliberately opened by the surgeon and is culture-positive or not cultured when the patient has at least one of the following symptoms: fever or localised pain or tenderness. An abscess, or other evidence of infection involving the deep incision is found on direct examination, during reoperation, or by histopathologic or radiologic examination. Diagnosis of a deep incisional SSI by a surgeon or attending physician Description of the intervention Dressings are widely used in the care of wounds. Several attributes of an ideal wound dressing have been described (BNF 2010; Goldman 1992; NICE 2008); these include: The ability of the dressing to absorb and contain exudate without leakage or strike-through. Lack of particulate contaminants left in the wound by the dressing. Thermal insulation. Impermeability to water and bacteria. Suitability of the dressing for use with different skin closures (sutures, staples). Avoidance of wound trauma on dressing removal. Frequency with which the dressing needs to be changed Provision of pain relief. Cosmesis and comfort. Effect on formation of scar tissue. Dressing products have evolved considerably in the last few decades, and now fall into broad, widely-recognised categories, namely: 1) the basic wound contact dressings such as gauze or cotton absorbents; 2) the advanced dressings such as hydrogels, hydrocolloids and films; and, 3) the anti-microbial and other specialist dressings. Within these groups there are many hundreds of dressing types available. For ease of comparison in this review, dressings have been classified into groups according to the British National Formulary (BNF) 2010 (BNF 2010), however, it is important to note that the distributors of dressings may vary from country to country, and that dressing names may also vary. For the purpose of this review, only the BNF groups that have a dressing referred to in the included studies have been detailed. 1. Basic wound contact dressings 1a. Absorbent dressings and surgical absorbents Absorbent dressings are applied directly to the wound. Surgical absorbents may be used as secondary absorbent layers in the management of heavily-exuding wounds. Examples include Primapore (Smith & Nephew), Mepore (Mölnlycke), and absorbent cotton gauze, BP b. Low adherence dressings and wound contact materials Low adherence dressings and wound contact materials are usually cotton pads that are placed directly in contact with the wound. They are either non-medicated (e.g. paraffin gauze dressing), or medicated (e.g. containing povidone iodine or chlorhexidine). Examples include paraffin gauze dressing, BP 1993, Xeroform Dressing - a non adherent petrolatum blend with 3% bismuth tribromophenate on fine mesh gauze. 3

6 2. Advanced dressings 2a. Vapour-permeable films Vapour-permeable films are permeable to water vapour and oxygen, but not to water or micro-organisms. They are normally transparent. Examples include OpSite (Smith & Nephew) and Tegaderm (3M). guideline reviewed the data from five trials relevant to this review, and concluded that current studies did not show convincing differences in dressing effectiveness in terms of reducing SSI (NICE 2008). Whilst the review methods were robust the search date was September 2007, thus studies published beyond this time have not been assessed. We are not aware of any other systematic reviews that assess the impact of dressings on infection in surgical wounds healing by primary intention. 2b. Hydrocolloid dressings Hydrocolloid dressings are occlusive dressings composed of a hydrocolloid matrix attached to a base (possibly film or foam). Fluid absorbed from the wound causes the hydrocolloid to liquefy. Examples include Comfeel (Coloplast) and DuoDerm (Conva- Tec, UK). O B J E C T I V E S To evaluate the effects of wound dressings versus no wound dressings or alternative dressings in preventing SSIs in surgical wounds healing by primary intention. 2c. Fibrous hydrocolloid dressing (Hydrofibre, spun hydrocolloid dressings) Fibrous hydrocolloid dressings are composed of sodium carboxymethylcellulose which gel when it comes into contact with fluid. Examples are Aquacel (ConvaTec, UK). M E T H O D S Criteria for considering studies for this review 2d. Polyurethane matrix hydrocolloid dressing Polyurethane matrix hydrocolloid dressings consist of two layers - a polyurethane gel matrix and a waterproof polyurethane top-film designed to act as a bacterial barrier. There is only one dressing of this type listed in the BNF: Cutinova Hydro (Smith & Nephew). How the intervention might work Current practice for surgical wounds healing by primary intention commonly involves placement of a dressing over the closed wound before the patient leaves the sterile environment of the operating theatre. This practice assumes that the risk of SSIs may be reduced by providing a barrier to environmental contamination. Furthermore, dressings may have additional roles in managing wound exudate, protecting wounds and their staples or sutures, and meeting patients expectations by hiding the wound, or, alternatively, facilitating health professionals observation of the wound. Why it is important to do this review Surgical wounds healing by primary intention are commonplace within all elective and emergency surgical practice. It is important to assess whether wound dressings have a potential role in reducing the risk of SSI. Such information can inform allocation of resources to appropriate treatments. Currently such decisions are made with limited review data. In the UK, a government funded Types of studies Randomised controlled trials (RCTs) that compared the application of wound dressings applied immediately postoperatively versus no wound dressing or alternative dressings to surgical wounds expected to heal by primary intention. Types of participants Studies involving adults or children (aged two years and over) who had undergone surgical procedures where healing of the surgical wound was planned by primary intention. Wounds of any contamination level (clean, clean/contaminated, contaminated and dirty) were eligible for inclusion. Procedures involving graft sites were excluded. Participants were required to have dressings applied in the operating theatre, immediately after closure of the skin. Studies where participants had infected wounds at the start of the study were excluded. Types of interventions The primary intervention was application of wound dressings that could be: Basic wound contact dressings, classed as surgical and nonsurgical absorbent dressings, low adherence dressings, impregnated/non impregnated gauze, and adhesive tape. Advanced wound dressings such as hydrogels, hydrocolloids and films. Antimicrobial and other specialist dressings. 4

7 We included comparisons of a dressing compared with no dressing (exposure), and comparisons of alternative dressings. We did not consider trials that compared different durations of the same dressing (timing trials), as these will form a separate review. Nor did we include trials where the application of topical gels or ointments to wounds (in the absence of a dressing comparator) were being evaluated, as we viewed these as different interventions. Types of outcome measures Primary outcomes Occurance of postoperative SSI as defined by the CDC criteria (Horan 2008), or the authors definition of SSI. We did not differentiate between superficial and deep-incisional infection. Secondary outcomes Cost: any measure of cost of treatment, or other aspects of resource use i.e. other equipment. Scarring: as reported by the author. Pain: reported using a validated scale or as reported by the author. Acceptability (patient and clinician): as reported by the author. Ease of removal (patient and clinician): as reported by the author. Studies were included in the review if they reported any of the specified outcomes. #5 MeSH descriptor Surgical Wound Dehiscence explode all trees #6 surg* NEAR/5 infect*:ti,ab,kw #7 surg* NEAR/5 wound*:ti,ab,kw #8 surg* NEAR/5 site*:ti,ab,kw #9 surg* NEAR/5 incision*:ti,ab,kw #10 (#4 OR #5 OR #6 OR #7 OR #8 OR #9) #11 (#3 AND #10) The search strategies for Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL can be found in Appendix 1; Appendix 2 and Appendix 3, respectively. The Ovid MEDLINE search was combined with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity- and precisionmaximizing version (2008 revision) (Lefebvre 2009). The Ovid EMBASE and EBSCO CINAHL searches were combined with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN) (SIGN 2010). There were no restrictions on the basis of date or language of publication. Searching other resources The bibliographies of all retrieved and relevant publications identified by these strategies were searched for further studies. While handsearches were not performed for this review, they are conducted by the Cochrane Wounds Group in order to inform the CENTRAL database, which we searched. We did not contact manufacturers regarding studies for inclusion. Search methods for identification of studies We searched the following electronic databases: Cochrane Wounds Group Specialised Register (searched 10 May 2011); The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2011, Issue 2); Ovid MEDLINE (1950 to April Week ); Ovid MEDLINE (In-Process & Other Non-Indexed Citations, May 9, 2011); Ovid EMBASE (1980 to 2011 Week 18); EBSCO CINAHL (1982 to 6 May 2011) We used the following search strategy in the Cochrane Central Register of Controlled Trials (CENTRAL): #1 MeSH descriptor Bandages explode all trees #2 (dressing* or hydrocolloid* or gauze* or hydrogel* or alginate* or bead or foam ):ti,ab,kw #3 (#1 OR #2) #4 MeSH descriptor Surgical Wound Infection explode all trees Data collection and analysis Selection of studies Two review authors independently assessed the studies titles and abstracts against the review s inclusion criteria. After this initial assessment, all studies that might potentially meet these criteria were obtained in full. Full papers were checked for eligibility by two review authors, with disagreements resolved by discussion and, where required, the input of a third review author. We extracted details of the eligible studies, and summarised them on a data extraction sheet. Two review authors extracted data independently. If data were missing from reports, attempts were made to contact the study authors to obtain the missing information. Studies that were published in duplicate were only included once, but the maximum amount of data were extracted from the papers. This process was followed by all review authors, with at least two review authors working independently. 5

8 Data extraction and management All data were extracted independently by two review authors. The following data were extracted: Country of origin of trial. Type of surgery. Classification of surgical contamination (see Table 1 for classification guide). Table 1. Classification of surgical contamination of included studies Classification Description Study Classification Clean only Non-infective operative wounds in which no inflammation is encountered, and neither the respiratory, alimentary, genitourinary tract nor the oro-pharyngeal cavity is entered. In addition these cases are elective, primarily closed, and drained with closed drainage system when required. Cosker 2005; De Win 1998; Lawrentschuk 2002; Law 1987; Michie 1994; Moshakis 1984; Vogt 2007; Wikblad 1995; Wynne 2004 Clean/ Contaminated only Contaminated only Dirty only Mixture Operative wounds in which respiratory, alimentary, genital or urinary tract is entered under controlled conditions and without unusual contamination. Specifically, operations involving the biliary tract, appendix, vagina and oropharynx are included in this category, provided no evidence of infection or a major break in sterile technique is encountered. Fresh, accidental wounds, operations with major breaks in sterile technique or gross spillage from the gastrointestinal tract, and incisions in which acute, non-purulent inflammation is encountered. Old traumatic wounds with retained devitalised tissue and those that involve existing clinical infection or perforated viscera. This definition suggests that organisms causing postoperative infection were present in the operative field before the operation. Persson 1995 Gardezi 1983;(clean, clean/contaminated and possibly contaminated) Hewlett 1996; (predominately clean, some clean/ contaminated and possibly contaminated) Holm 1998; (clean, clean/contaminated and contaminated) Phan 1993; (clean, clean/contaminated) Shinohara 2008; (clean, clean/contaminated and possibly contaminated) 6

9 Table 1. Classification of surgical contamination of included studies (Continued) No classification Rohde 1979 Eligibility criteria and baseline participant data. Details of the dressing/treatment regimen received by each group, including the duration that the dressing was in situ. Primary and secondary outcome(s) (with definitions). Outcome data for primary and secondary outcomes (by group). Duration of follow-up. Number of withdrawals (by group). Assessment of risk of bias in included studies Two review authors independently assessed each included study. Assessment was undertaken using the Cochrane Collaboration tool for assessing risk of bias (Higgins 2009). The risk of bias tool considers six domains: sequence generation, allocation concealment, blinding, incomplete outcome data, freedom from selective reporting, and other issues (i.e. serious baseline imbalance). A risk of bias table was completed for each eligible study; these data were combined into a risk of bias summary figure where judgements for each domain are tabulated by study. Measures of treatment effect Results were presented with 95% confidence intervals (CI). Estimates for dichotomous outcomes (e.g. infected: yes/no) were reported as risk ratios (RR) (Deeks 2002). Continuous data (e.g. pain) were reported as mean differences (MD) and we calculated overall effect sizes (with 95% CI). Dealing with missing data This issue was not considered in the protocol. The problem of missing data is common in trials - especially those of poor quality. Excluding participants from the analysis after randomisation, or ignoring participants lost to follow-up can, in effect, undo the process of randomisation, and thus, potentially, introduce bias into the trial. For our primary outcome, SSI, we assumed that where randomised participants were not included in an analysis, they did not have an SSI (that is they were considered in the denominator but not the numerator). Given the relatively small number of SSI events anticipated, this seemed the most appropriate assumption. Where a trial did not specify participant group numbers prior to dropout, only complete case data were presented. Data for all secondary outcomes were presented as complete case analysis. Data synthesis Details of included studies were combined in narrative review according to dressing type and stratified by surgical contamination level. Both clinical and statistical heterogeneity were explored. Where appropriate, data were pooled using meta-analysis (conducted using RevMan 5), that is, where studies appeared similar in terms of level of surgical contamination, intervention type and duration, and outcomes. Statistical heterogeneity was assessed using the chi-squared test (a significance level of P value less than 0.1 was considered to indicate heterogeneity), and the I² test (Higgins 2003). The I² test examines the percentage of total variation across studies due to heterogeneity rather than to chance. Values of I² over 50% may represent substantial heterogeneity (Higgins 2009). In the absence of clinical heterogeneity, and in the presence of statistical heterogeneity (I² over 50%), a random-effects model was used. Where there was no clinical or statistical heterogeneity, a fixed-effect model was applied. R E S U L T S Description of studies See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification. See Characteristics of included studies; Characteristics of excluded studies for full details of studies identified. Two studies are awaiting classification: one requires translation (Pizarro 2001) and we are uncertain if one study is an RCT (Marinovic author contacted). We are not aware of any relevant on-going studies (checked ISRCTN register 7th May 2011). Included studies A total of 16 RCTs met the inclusion criteria. There were 12 twoarm trials, and four three-arm trials; in one of the latter, two of the three groups were randomised to receive different brands of a film dressing (Cosker 2005). For this review, these two film-dressing groups were combined into one group and the trial was treated as a two-arm trial. This was not undertaken for the three remaining three-arm trials, since all groups were deemed to be receiving different interventions, and so all relevant comparisons were included. In all trials the surgical procedure took place in a hospital operating theatre. Only five (31%) of the included trials had been 7

10 published in the last 10 years (Cosker 2005; Lawrentschuk 2002; Shinohara 2008; Vogt 2007; Wynne 2004). The remaining trials were between 12 and 31 years old. The trials took place in several different countries: 11 were conducted in Europe, four in the UK (Cosker 2005; Hewlett 1996; Law 1987; Moshakis 1984); two in Belgium (De Win 1998; Phan 1993); two in Sweden (Persson 1995; Wikblad 1995); two in Denmark (Holm 1998; Vogt 2007); and one in Germany, (Rohde 1979). Two of the remaining five trials were conducted in Australia (Lawrentschuk 2002; Wynne 2004); one in Pakistan (Gardezi 1983); one in the USA (Michie 1994); and one in Japan (Shinohara 2008). One trial was published in German, and a translation was required (Rohde 1979). The types of surgical procedures undertaken were varied and included cardiac and/or vascular surgery (Shinohara 2008; Vogt 2007; Wikblad 1995; Wynne 2004), abdominal surgery and/or gastrointestinal surgery (Holm 1998; Persson 1995; Rohde 1979), or a number of different surgical procedures within the same trial (Gardezi 1983; Hewlett 1996). The surgical procedures in each trial were classified as having been clean, clean/contaminated, contaminated or dirty, or a combination of these (see Table 1 for classification). When the type of surgery performed was unclear, this was recorded (Rohde 1979). Studies also compared a range of different dressing types and regimens as described below and in Table 2. Table 2. Overview of dressing comparisons Basic wound contact dressings Advanced Other Exposed Timing Study (contamination class) Surgical absorbent Adhesive Tape Low adherence Hydrocolloid Films Other Polyurethanedressing matrix hydrocolloid Exposed Early removal of dressing Cosker 2005 (Clean) Primapore (duration not clear) Tegaderm (duration not clear) Opsite (duration not clear) De Win 1998 (Clean) Mepore (duration not clear) Tegaderm (duration not clear) Gardezi 1983 (Mixed) Hewlett 1996 (Mixed) Gauze (changed after 48 h) Mepore Polyurethane membrane (left in situ until suture removal) Opsite Holm 1998 (Mixed) Comfeel (removed day Mepore (removed at 2 8

11 Table 2. Overview of dressing comparisons (Continued) 10 after surgery) days after surgery) Law 1987 (Clean) Gauze (removed day 5 after surgery) Opsite (removed day 5 after surgery) Wound left uncovered (unless discharging) Interpose Lawrentschuk (removed days after (Clean) surgery) Paraffin tulle gras (removed 5 days after surgery) Michie 1994 (Clean) Xeroform (removed 7-10 days after surgery) Duo- Derm (removed 7-10 days after surgery) Moshakis 1984 (Clean) Gauze secured with Transpore or Elastoplast (in place at hospital discharge) Tegaderm (in place at hospital discharged) Persson 1995 (Clean/ contaminated) DuoDerm E (left in place until hospital discharge, or infection occurred) Absorbent dressing (removed morning after surgery) Phan 1993 (Clean) Gauze dressing changed twice daily Vaseline ointment removed and re-applied twice daily 9

12 Table 2. Overview of dressing comparisons (Continued) Rohde 1979 (Unclear) Fixomull (duration unknown) Opsite (duration unknown) Shinohara 2008 (Mixed) Gauze (removed day 7 after surgery) Karayahesive (removed day 7 after surgery) Vogt 2007 (Clean) Mepore (removed day 4 after surgery) Aquacel (removed day 4 after surgery) Wikblad 1995 (Clean) Absorbent dressing (removed 5 days after surgery) DuoDerm (removed 5 days after surgery) Cutinova Hydro (removed 5 days after surgery) Wynne 2004 (Clean) DuoDerm Thin (removed 5 days after surgery) Opsite (removed 5 days after surgery) Primapore (removed day 2 after surgery) Excluded studies In total, 42 studies were excluded after screening of the full text. There were a number of reasons for exclusions including nine studies that were not RCTs and six studies that included wounds healing by secondary intention i.e. wounds were left open, or had broken open, and were healing from the bottom-up. Full details are given in the Characteristics of excluded studies. Risk of bias in included studies We classified trials as being at high risk of bias if they were rated no for any of three key criteria (i.e. randomisation sequence, allocation concealment, incomplete outcome data). In total, three trials were deemed to be at high risk of bias (see Figure 1 for risk of bias summary). Indeed, on the basis of our assessment we could not be confident that any trials were at low risk of bias. 10

13 Figure 1. Methodological quality summary: review authors judgements about each methodological quality item for each included study. 11

14 Allocation Adequacy of randomisation process All included studies were described as randomised, however, only three provided information that confirmed that adequate sequence generation had taken place. Lawrentschuk 2002 randomised participants at the time of skin closure using a computer-generated sequence placed in an envelope (no further details about this process are provided). Michie 1994 used a computer-generated randomised table with blocks of four to determine which dressing was used on each side of the wound (patients acting as their own control). Finally, we also considered there was adequate sequence generation for Vogt 2007, as prior to the start of the study an uninvolved person mixed 160 notes, half of them marked with one dressing type and half marked with the other, placing them in consecutive marked envelopes. We did not consider that the randomisation process used by Moshakis 1984 could be classed as adequate, since some participants acted as their own controls and some did not. We also felt that the process adopted by Wikblad 1995, where a secretary randomly selected a number from one to three (a number of each dressing type) and put the number on the anaesthesiologist s order sheet, could not be considered to be random given the limited information provided. Allocation concealment In 15 of the 16 included trials it was unclear whether the randomisation sequence was concealed at the point of participant contact. Many of the studies were published prior to the development of the CONSORT statement, and the reports lacked methodological details. We classed Vogt 2007 as being at low risk of bias for allocation concealment as the report stated, In the operating theatre the envelope was opened and the relevant dressing applied to the wound. The randomisation method used by Wikblad 1995, described above, was not classed as adequate for this criterion. Blinding Whilst it could be difficult to blind clinicians and participants to the dressing type received, and as the diagnosis of SSI can be subjective, we were particularly interested to see if any trial had conducted blinded outcome assessment (of SSI), i.e. assessment by an independent assessor, unaware of participant allocation. Information about blinding, however, was seldom reported, and only Wynne 2004 confirmed that no blinding had taken place. Accordingly, the remaining 15 studies were recorded as unclear. Incomplete outcome data In a number of trials participants were excluded from the analysis in large enough numbers to threaten bias. Holm 1998 excluded 23 of 73 randomised participants from the analysis; Persson 1995 excluded seven of 62 randomised participants; Phan 1993 excluded 26 of 207 randomised participants; Vogt 2007 excluded 24 of 160 randomised participants; and, in Wikblad 1995, 34 of 250 randomised participants were lost to follow-up in the first week. No trial implemented any methods to investigate or address the impact of missing data. Other potential sources of bias Six trials were recorded as having other sources of bias. De Win 1998 and Moshakis 1984 included multiple wounds per patient, which did not seem to be accounted for in the analysis. Holm 1998, Persson 1995 and Wynne 2004 also varied timing - as well as dressing type - within the trial arms. It was difficult to judge whether or not these trials were simply pragmatic ones, assessing each dressing as it was used normally in that setting. Consequently, they were classed as unclear in terms of bias. Finally, Cosker 2005 appeared to have baseline imbalance in age, although limited data were reported. Effects of interventions Dressings compared with exposed wounds Comparion 1: Basic wound contact dressings compared with wound exposure (2 trials; 319 participants) Two trials, involving a total of 319 participants, compared basic wound contact dressings with wound exposure. Law 1987 was a three-arm trial where the surgical procedure had a clean contamination classification (112 participants involved in this comparison). The trial compared a basic wound contact dressing (removed after five days and changed if wound was discharging), with exposed wounds. Phan 1993 undertook a surgical procedure with clean and clean/contaminated contamination classification and compared a basic wound contact dressing (changed twice daily) with exposed wounds that were treated with petroleum jelly (Vaseline). Primary outcome: SSI 12

15 Clean surgery Law 1987: there was no statistically significant difference in the number of SSIs in the exposed group (1/53; 2%) compared with the basic wound contact-dressed group (3/59; 5%) (RR for developing SSI in the basic wound contact-dressed group compared with the exposed group = 2.69; 95% CI 0.29 to 25.13) (Analysis 1.1). As only four patients (out of 170) were reported as being lost to follow-up (groups not reported), a complete case analysis is presented. Other surgery types Phan 1993: there was no statistically significant difference in the number of SSIs in the exposed (ointment) group (29/105; 28%) compared with basic wound contact-dressed group (21/102; 21%) (RR for developing an SSI in the basic wound contact-dressed group compared with the exposed group = 0.75; 95% CI 0.46 to 1.22) (Analysis 1.1).Twenty-six participants from this trial (13% of the total) were excluded after randomisation (in our analysis these participants were included in the denominator but not numerator, i.e. it was assumed that they did not have an SSI, as per our protocol regarding the treatment of missing data). This study was classed as being at a high risk of bias. We decided not to pool the data from these trials as the surgery undertaken in Phan 1993 was classed as clean and clean/contaminated, and an ointment was applied, whereas the surgery undertaken in Law 1987 was clean and wounds were not treated with an ointment. Secondary outcomes There is insufficient evidence regarding whether wound coverage with basic wound contact dressings (versus leaving a wound exposed) affects the rate of SSIs surgical wounds that are healing by primary intention; what evidence there is indicates no effect. The use of dressings incurs additional costs, but there are no cost-effectiveness data available from these studies to facilitate informed decision making. Existing evidence is both sparse and poor quality. Comparison 2: Advanced dressings compared with exposed wounds (1 trial; 107 participants) This was a second comparison of the three-arm study by Law 1987 that compared a film-dressed wound with exposed wounds in 107 participants. Primary outcome: SSI. Clean surgery Law 1987: there was no statistically significant difference in the number of SSIs in the group left with wounds exposed (1/53; 2%) compared with the film-dressed group (5/54; 9%) (RR for developing an SSI in the film-dressed group compared to the exposed group = 4.91; 95% CI 0.59 to 40.61) (Analysis 2.1). Secondary outcomes Clean surgery Law 1987: reported that the total dressing costs for the basic wound contact-dressed group were GBP 6.60 compared to GBP 0.80 in the exposed group. No detailed information was presented i.e. the cost of complications, duration of stay in hospital and nurse time. The cost of dressing data alone is of limited value to decision makers. Law 1987 also report that there was no difference in quality of final scar between the exposed group and the basic wound contact-dressed group, but no data were presented, nor was any information provided regarding who measured this outcome, how it was measured, and how long after surgery. Finally, the study reported no difference in dressing preference as measured on a linear analogue scale. No further information or data were presented. Other surgery types Phan 1993 did not present data on secondary outcomes. Summary: Basic wound contact dressings compared with exposed wounds Clean surgery Law 1987: total dressing costs for the film group were GBP compared to GBP 0.80 in the exposed group. Summary: Advanced wound contact dressings compared with exposed wound There is no evidence that using a film dressing or leaving a wound exposed leads to a difference in the number of SSI in surgical wounds healing by primary intention. Law 1987 reported that film dressings were more costly per unit. Dressings compared with dressings Comparison 3: Comparisons between different basic wound contact dressings (1 trial; 50 participants) Lawrentschuk 2002 undertook a surgical procedure with a clean contamination classification and compared a paraffin tulle dressing with a non-adherent dressing in 50 participants (25 in each arm). 13

16 Both dressing types were applied in the same way. In both groups a compressible, combined dressing was placed over the evaluated dressings with an adhesive elastic dressing then placed over these. Primary outcome: SSI Clean surgery Lawrentschuk 2002: There was no statistically significant difference in the number of SSIs in the non-adherent dressing group (3/25; 12%) compared with the paraffin tulle dressing group (0/ 25; 0%) (RR 0.14; 95% CI 0.01 to 2.63). (Analysis 3.1). No relevant secondary outcomes were reported. Summary: Basic wound contact dressings compared with different basic wound contact dressings There is no evidence to suggest that using a non-adherent dressing or a paraffin tulle dressing leads to a difference in the number of SSIs in surgical wounds healing by primary intention. an interim analysis. It is important to note that some participants had multiple wounds. It was not clear from the trial report whether one, or more wounds, were followed-up for each participant. Law 1987: there was no statistically significant difference in the number of SSIs in the basic wound contact-dressed group (3/59; 5%), compared with the film-dressed group (5/54; 9%) (RR 1.82, 95% CI 0.46 to 7.26) (Analysis 4.1). Moshakis 1984: did not report SSI data. Wynne 2004: there was no statistically significant difference in the number of SSIs in the basic wound contact-dressed group (6/243; 3%), compared with the film-dressed group (9/227; 4%) (RR % CI 0.58 to 4.44) (Analysis 4.1). It was decided that data from four trials with participants undergoing clean surgery could be pooled (Cosker 2005; De Win 1998; Law 1987; Wynne 2004) (Analysis 4.2). Whilst Law 1987 and Wynne 2004 were three-arm trials, this was the only meta-analysis conducted with their data, so the complete groups relevant to this pooling were used. De Win 1998 reported zero SSI events. Moshakis 1984 did not report SSI data. Pooled data showed there was no significant difference in the number of SSIs between basic wound contact-dressed groups compared with film-dressed groups (RR 1.34; 95% CI 0.70 to 2.55) (Analysis 4.2). Comparison 4: Basic wound contact dressings compared with film dressings (8 trials; 1087 participants) Eight trials compared a basic wound contact dressing with a film dressing. Five of these trials evaluated wounds resulting from clean surgical procedures (Cosker 2005; De Win 1998; Law 1987; Moshakis 1984; Wynne 2004) and three evaluated wounds resulting from surgical procedures with mixed, or unclear, contamination classifications (Gardezi 1983; Hewlett 1996; Rohde 1979). The trials included a variety of basic wound contact dressings including gauze and surgical absorbents. Similarly, whilst the comparators were all film dressings, different brands were evaluated (five trials evaluated Opsite (Smith & Nephew), three Tegaderm (3M Healthcare), and one an unnamed brand (Cosker 2005 evaluated two film dressings). Primary outcome: SSI Clean surgery Cosker 2005: there was no statistically significant difference in the number of SSIs in the basic wound contact-dressed group (5/ 100; 5%), compared with the film-dressed group (9/200; 5%) (RR 0.90; 95% CI 0.31 to 2.61) (Analysis 4.1). De Win 1998: there was no statistically significant difference in the number of SSIs in this trial. No SSIs developed in the basic wound contact-dressed group (n = 6) or the film-dressed group (n = 8). This was a small, industry-funded study, which was presented as Other surgery types (including unclear) Gardezi 1983: conducted several surgical procedures that were classified as clean, clean/contaminated and possibly contaminated. There was no statistically significant difference in the number of SSIs in the basic wound contact-dressed group (6/50; 12%) compared with the film-dressed group (3/50; 6%) (RR % CI 0.13 to 1.89) (Analysis 4.1). Hewlett 1996: did not report SSI data. Rohde 1979: was classed as unclear in terms of surgical procedure(s) undertaken. In total, 24/46 (52%) of participants in the basic wound contact dressing group had a mild wound infection compared with 14/44 (32%) in the film-dressed group (RR 0.61; 95% CI 0.36 to 1.02 in favour of the film dressing preventing SSI). This estimate was not statistically significant, (Analysis 4.1), although it may have been underpowered. Additionally, data for this paper were taken from a translation, and it was not possible to assess the quality of the trial. It is, however, over 30 years old, and potentially not a robust, single source of data on which to base decisions regarding dressing use. Given the difficulty in classifying the type of surgical procedure(s) undertaken in Rohde 1979 we did not pool this study with Gardezi However, since it is not clear what impact the level of surgical contamination might have on the potential action of dressings in SSI prevention, in a sensitivity analysis we pooled all six trials that reported SSI data in this comparison, regardless of contamination level (Cosker 2005; De Win 1998; Gardezi 1983; Law 1987; Rohde 1979; Wynne 2004). Pooled data showed there was no significant difference in the number of SSIs between basic wound 14

17 contact-dressed groups compared with film-dressed groups (RR 0.84; 95% CI 0.58 to 1.24) (Analysis 4.1). Secondary outcomes the basic wound contact-dressed group and DEM 3.60 in the film group. Rohde 1979 also reported figures for ease of dressing removal and pain, but as there was no information about how these data were obtained, it is difficult to interpret them and they are not presented here (see Characteristics of included studies). Clean surgery De Win 1998: reported the total mean cost of dressings for each group in Belgian Francs (now replaced by the Euro) - BEL 11.5 in the basic wound contact-dressed group and BEL 14.3 in the film-dressed group. No further economic data were presented in this study. Law 1987 reported the mean cost of the dressings in each group: GBP 6.60 in the basic wound contact group and GBP in the film group. Moshakis 1984: reported the levels of pain in each group. This was measured using a patient-assessed linear scale (1 to 10) where one corresponded to no discomfort and 10 to extremely uncomfortable or painful. Mean pain levels in the basic wound contact group were 5.1 (SD 2.78) compared with 1.6 (SD 1.48) in the filmdressed group, a statistically significant finding in favour of film dressings: mean difference (MD -3.50; 95% CI to -2.71) (Analysis 4.3). Participants and treating nurses were also asked to rate their acceptability of the dressings, which again were measured using a linear scale where one corresponded to no trouble and 10 to very troublesome. The mean response of basic wound contactdressed participants was 4.2 (SD 2.46) and the mean response of the film-dressed group 1.3 (SD 1.17) (MD -2.90; 95% CI to -2.21, again favouring the film-dressed group) (Analysis 4.4). The mean acceptability response of the treating nurse was 5.4 (SD unknown) in the basic wound contact group and 1.2 (SD unknown) in the film group. However, all analyses in this trial must be interpreted with caution, as allocation concealment was not adequate; additionally there was no indication in the paper that any statistical procedures had been used to deal with the potential violation of assumptions caused by the inclusion of some participants as their own controls (bilateral excisions). This study was deemed to be at high risk of bias. Other surgery types (including unclear) Gardezi 1983: reported a measure for pain in each group, no details were provided regarding collection of these data or how they could be interpreted. Hewlett 1996: reported the cost of dressings (including and excluding procedure pack) as GBP 1.60 for the basic wound contact-dressed group compared with GBP 1.46 for the film-dressed group or GBP 4.36 compared with GBP 2.84 when procedure packs were included. Rohde 1979 reported the cost per patient in Deutsch Marks (now replaced by the Euro) as DEM in Summary: Basic wound contact dressings compared with film dressings There is no evidence for a difference in the number of SSIs in surgical wounds healing by primary intention when treated with a basic wound contact dressing or a film dressing. Rohde 1979 suggested an increase in the proportion of SSIs developed in basic wound contact-dressed wounds compared with film-dressed wounds (52% compared with 32%) but this was not statistically significant. One trial, Moshakis 1984 suggested that people with wounds treated with film dressings reported less pain and the dressing were more acceptable to the participants. Once again, however, these results should be interpreted with caution, since the trial was deemed to be at high risk of bias. The cost data presented were too limited to allow any conclusions based on costs versus benefits to be assessed. Comparison 5: Basic contract wound dressings compared with hydrocolloid dressings (6 trials; 792 participants) Six trials investigated the effect of a basic wound contact dressing compared with a hydrocolloid dressing. The basic wound contact dressings were predominantly gauze. Four trials investigated the same brand of hydrocolloid dressing (DuoDerm, ConvaTec UK). Primary outcome: SSI Clean surgery Michie 1994: there was no statistically significant difference in the number of SSIs in this small trial, in which the surgical procedure was given a clean contamination classification. None of the 28 wound halves randomised to the basic wound contact dressing or the hydrocolloid dressing developed an SSI. Wikblad 1995: no clear data were presented for SSI. The authors reported that 11 participants were treated with antibiotics postoperatively, and eight of these participants had infections in the sternum (of whom five were in the basic wound contact dressing group). No further information was provided. This study was classed as being at a high risk of bias due to a large amount of missing data. Wynne 2004: the results of this pragmatic trial reflect the impact of varying both dressing type and the time it was in situ between 15