When planning care for a burn patient, the nurse would anticipate applying which topical medication?

Cochrane Database Syst Rev. 2017 Jul; 2017(7): CD011821.

Monitoring Editor: Gill Norman,

University of Manchester, Manchester Academic Health Science Centre, Division of Nursing, Midwifery & Social Work, School of Health Sciences, Faculty of Biology, Medicine & Health, Jean McFarlane Building, Oxford Road, ManchesterUK, M13 9PL

National Institute for Health and Care Excellence (NICE), Evidence Information Services, Level 1A, City Tower, Piccadilly Plaza, ManchesterUK, M1 4BT

University Hospital of South Manchester NHS Foundation Trust, Wythenshawe Hospital, Burn Centre, Acute Block, Southmoor Road, ManchesterUK, M23 9LT

Jawaharlal Institute of Postgraduate Medical Education and Research (JIPMER), Plastic Surgery, Department of Plastic Surgery, 4th Floor, Superspeciality Block, JIPMER, PuducherryPuducherryIndia, 605006

University Hospital of South Manchester NHS Foundation Trust, Wythenshawe Hospital, Microbiology, Southmoor Road, ManchesterUK, M23 9LT

Abstract

Background

Burn wounds cause high levels of morbidity and mortality worldwide. People with burns are particularly vulnerable to infections; over 75% of all burn deaths (after initial resuscitation) result from infection. Antiseptics are topical agents that act to prevent growth of micro‐organisms. A wide range are used with the intention of preventing infection and promoting healing of burn wounds.

Objectives

To assess the effects and safety of antiseptics for the treatment of burns in any care setting.

Search methods

In September 2016 we searched the Cochrane Wounds Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL), Ovid MEDLINE, Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations), Ovid Embase, and EBSCO CINAHL. We also searched three clinical trials registries and references of included studies and relevant systematic reviews. There were no restrictions based on language, date of publication or study setting.

Selection criteria

We included randomised controlled trials (RCTs) that enrolled people with any burn wound and assessed the use of a topical treatment with antiseptic properties.

Data collection and analysis

Two review authors independently performed study selection, risk of bias assessment and data extraction.

Main results

We included 56 RCTs with 5807 randomised participants. Almost all trials had poorly reported methodology, meaning that it is unclear whether they were at high risk of bias. In many cases the primary review outcomes, wound healing and infection, were not reported, or were reported incompletely.

Most trials enrolled people with recent burns, described as second‐degree and less than 40% of total body surface area; most participants were adults. Antiseptic agents assessed were: silver‐based, honey, Aloe Vera, iodine‐based, chlorhexidine or polyhexanide (biguanides), sodium hypochlorite, merbromin, ethacridine lactate, cerium nitrate and Arnebia euchroma. Most studies compared antiseptic with a topical antibiotic, primarily silver sulfadiazine (SSD); others compared antiseptic with a non‐antibacterial treatment or another antiseptic. Most evidence was assessed as low or very low certainty, often because of imprecision resulting from few participants, low event rates, or both, often in single studies.

Antiseptics versus topical antibiotics

Compared with the topical antibiotic, SSD, there is low certainty evidence that, on average, there is no clear difference in the hazard of healing (chance of healing over time), between silver‐based antiseptics and SSD (HR 1.25, 95% CI 0.94 to 1.67; I2 = 0%; 3 studies; 259 participants); silver‐based antiseptics may, on average, increase the number of healing events over 21 or 28 days' follow‐up (RR 1.17 95% CI 1.00 to 1.37; I2 = 45%; 5 studies; 408 participants) and may, on average, reduce mean time to healing (difference in means ‐3.33 days; 95% CI ‐4.96 to ‐1.70; I2 = 87%; 10 studies; 979 participants).

There is moderate certainty evidence that, on average, burns treated with honey are probably more likely to heal over time compared with topical antibiotics (HR 2.45, 95% CI 1.71 to 3.52; I2 = 66%; 5 studies; 140 participants).

There is low certainty evidence from single trials that sodium hypochlorite may, on average, slightly reduce mean time to healing compared with SSD (difference in means ‐2.10 days, 95% CI ‐3.87 to ‐0.33, 10 participants (20 burns)) as may merbromin compared with zinc sulfadiazine (difference in means ‐3.48 days, 95% CI ‐6.85 to ‐0.11, 50 relevant participants). Other comparisons with low or very low certainty evidence did not find clear differences between groups.

Most comparisons did not report data on infection. Based on the available data we cannot be certain if antiseptic treatments increase or reduce the risk of infection compared with topical antibiotics (very low certainty evidence).

Antiseptics versus alternative antiseptics

There may be some reduction in mean time to healing for wounds treated with povidone iodine compared with chlorhexidine (MD ‐2.21 days, 95% CI 0.34 to 4.08). Other evidence showed no clear differences and is of low or very low certainty.

Antiseptics versus non‐antibacterial comparators

We found high certainty evidence that treating burns with honey, on average, reduced mean times to healing in comparison with non‐antibacterial treatments (difference in means ‐5.3 days, 95% CI ‐6.30 to ‐4.34; I2 = 71%; 4 studies; 1156 participants) but this comparison included some unconventional treatments such as amniotic membrane and potato peel. There is moderate certainty evidence that honey probably also increases the likelihood of wounds healing over time compared to unconventional anti‐bacterial treatments (HR 2.86, 95% C 1.60 to 5.11; I2 = 50%; 2 studies; 154 participants).

There is moderate certainty evidence that, on average, burns treated with nanocrystalline silver dressings probably have a slightly shorter mean time to healing than those treated with Vaseline gauze (difference in means ‐3.49 days, 95% CI ‐4.46 to ‐2.52; I2 = 0%; 2 studies, 204 participants), but low certainty evidence that there may be little or no difference in numbers of healing events at 14 days between burns treated with silver xenograft or paraffin gauze (RR 1.13, 95% CI 0.59 to 2.16 1 study; 32 participants). Other comparisons represented low or very low certainty evidence.

It is uncertain whether infection rates in burns treated with either silver‐based antiseptics or honey differ compared with non‐antimicrobial treatments (very low certainty evidence). There is probably no difference in infection rates between an iodine‐based treatment compared with moist exposed burn ointment (moderate certainty evidence). It is also uncertain whether infection rates differ for SSD plus cerium nitrate, compared with SSD alone (low certainty evidence).

Mortality was low where reported. Most comparisons provided low certainty evidence that there may be little or no difference between many treatments. There may be fewer deaths in groups treated with cerium nitrate plus SSD compared with SSD alone (RR 0.22, 95% CI 0.05 to 0.99; I2 = 0%, 2 studies, 214 participants) (low certainty evidence).

Authors' conclusions

It was often uncertain whether antiseptics were associated with any difference in healing, infections, or other outcomes. Where there is moderate or high certainty evidence, decision makers need to consider the applicability of the evidence from the comparison to their patients. Reporting was poor, to the extent that we are not confident that most trials are free from risk of bias.

Plain language summary

Antiseptics for Burns

Review question

We reviewed the evidence about whether antiseptics are safe and effective for treating burn wounds.

Background

Burn wounds cause many injuries and deaths worldwide. People with burn wounds are especially vulnerable to infections. Antiseptics prevent the growth of micro‐organisms such as bacteria. They can be applied to burn wounds in dressings or washes, which may help to prevent infection and encourage wound healing. We wanted to find out if antiseptics are more effective than other types of treatment, or whether one antiseptic may be more effective than others, in reducing infection and speeding up healing.

Study characteristics

In September 2016 we searched for randomised controlled trials (RCTs) involving antiseptic treatments for burn wounds. We included 56 studies with 5807 participants. Most participants were adults with recent second‐degree burns taking up less than 40% of their total body surface area. The antiseptics used included: silver‐based, honey, iodine‐based, chlorhexidine or polyhexanide (biguanides). Most studies compared antiseptics with a topical antibiotic (applied to the skin). A smaller number of studies compared antiseptics with a non‐antibacterial treatment, or with another antiseptic.

Key results

The majority of studies compared antiseptic treatments with silver sulfadiazine (SSD), a topical antibiotic used commonly in the treatment of burns. There is low certainty evidence that some antiseptics may speed up average times to healing compared with SSD. There is also moderate certainty evidence that burns treated with honey probably heal more quickly compared with those treated with topical antibiotics. Most other comparisons did not show a clear difference between antiseptics and antibiotics.

There is evidence that burns treated with honey heal more quickly (high certainty evidence) and are more likely to heal (moderate certainty evidence) compared with those given a range of non‐antibacterial treatments, some of which were unconventional. Burns treated with antiseptics such as nanocrystalline silver or merbromin may heal more quickly on average than those treated with Vaseline gauze or other non‐antibacterial treatments (moderate or low certainty evidence). Comparisons of two different antiseptics were limited but average time to healing may be slightly quicker for wounds treated with povidone iodine compared with chlorhexidine (low certainty evidence). Few participants in the studies experienced serious side effects, but this was not always reported. The results do not allow us to be certain about differences in infection rates. Mortality was low where reported.

Quality of the evidence

Most studies were not well reported and this makes it difficult to be sure if they were at risk of bias. In many cases a single (often small) study provides all the evidence for the comparative effects of the different treatments; and some similar studies provided conflicting results. Where there is moderate or high certainty evidence clinicians will need to consider whether the evidence from the comparison is relevant to their patients.

This plain language summary is up to date as of September 2016.

Summary of findings

Background

Description of the condition

A burn can be defined as an injury to the skin or other organic tissue caused by thermal trauma (Hendon 2002). Burns are caused by heat (including contact with flames, high temperature solids (contact burns) and liquids (scalds)), chemicals, electricity, friction or abrasion, and radiation (including sunburn and radioactivity). Respiratory damage, as a consequence of smoke inhalation, is also considered a type of burn (Hendon 2002).

Incidence and impact

Burn injuries are a considerable source of morbidity and mortality (Mock 2008). As outlined by the World Health Organization (WHO), the burden of injury falls predominantly on people living in low‐ and middle‐income countries; over 95% of the 300,000 annual deaths from fires occur in these countries (Mock 2008). Total burn mortality is inversely correlated with both national income and income inequality (Peck 2013). The much greater number of injuries resulting in disability and disfigurement are also disproportionately concentrated in low‐ and middle‐ income countries (Mock 2008). Fire‐related burns have been estimated to account for 10 million lost disability‐adjusted life years (DALYs) every year (WHO 2002), a figure that does not include the social and personal impact of non‐disabling disfigurement.

Additional mortality and morbidity are caused by other types of burns including scalding, and electrical and chemical burns (American Burn Association 2013). Globally, children and young people, and women are disproportionately affected by burn injuries, while the types and causes of injury in children differ somewhat from those seen in adults (Peck 2012).

Although, both incidence of burns and associated morbidity and mortality are much lower in high‐income countries, they are nevertheless significant. Annually in the UK around 250,000 people suffer a burn; 175,000 attend a hospital emergency department with a burn and, of these, approximately 13,000 are admitted to hospital and 300 die (National Burn Care Review 2001). In the USA, the figures for those receiving medical treatment were 450,000 with 40,000 hospitalisations and 3400 deaths (American Burn Association 2013). These data indicated that, in contrast to the global pattern, a majority of people with burns were male (69%), and while children aged under five years accounted for 20% of all cases, 12% were people aged 60 years or older (American Burn Association 2013).

Burn severity and extent

The severity of burns is categorised by the depth of the tissues affected; in the case of burns to the skin, this is the layers of cells in the skin (Demling 2005). Epidermal burns (sometimes known as first‐degree burns) are confined to the epidermis (outer surface of the skin), are not usually significant injuries, and heal rapidly and spontaneously. Partial‐thickness burns (sometimes known as second‐degree burns) involve varying amounts of the dermis (skin) and may become deeper and heal with varying amounts of scarring, which will be determined partly by the depth of the burn. Partial‐thickness burns are divided into superficial and deep partial‐thickness wounds: superficial partial‐thickness burns extend into the papillary or superficial upper layer of the dermis, whilst deep partial‐thickness burns extend downward into the reticular (lower) layer of the dermis. Full‐thickness burns (sometimes known as third‐degree burns) extend through all the layers of the skin. Where full‐thickness burns extend beneath the skin layers, into underlying structures (fat, muscle, bone); they are sometimes called full‐thickness and/or fourth‐degree burns) (Demling 2005; European Practice Guidelines 2002).

The age of people with burns affects their prognosis, with infants and older people having poorer outcomes (Alp 2012; DeSanti 2005). The area of a burn will also be key to the time taken to heal, and also to the risk of infection (Alp 2012). Burn size is determined by the percentage of the total body surface area that is burned; estimating this can be difficult, particularly in children; the most accurate method uses the Lund and Browder chart (Hettiaratchy 2004).

The depth of burn and its location may be predictors of psychological, social, and physical functioning following treatment (Baker 1996). Most extensive burns are a mixture of different depths, and burn depth can change and increase in the acute phase after the initial injury; the extent to which this occurs will depend on the effectiveness of the initial treatment (resuscitation) (Hettiaratchy 2004).

Burn wound infection

Infections are a potentially serious complication in people with burns. US data indicated that over a 10‐year period more than 19,000 complications in people with burns were reported. While 31% of these were recorded as pulmonary complications, 17% were wound infections, or cellulitis, or both, and 15% were recorded as septicaemia (a serious, life‐threatening illness caused by bacteria in the bloodstream) or other infectious complications (Latenser 2007). We were unable to locate other large‐scale international data for infection‐related complication rates.

Up to 75% of all burn deaths following initial resuscitation are a consequence of infection rather than more proximal causes such as osmotic shock and hypovolaemia (types of changes in the concentration of fluids in the body) (Bang 2002; Fitzwater 2003). Although this figure includes other types of hospital/healthcare‐acquired infections such as pneumonia, a substantial proportion follow an infection which would meet accepted criteria for infections of burn wounds (Alp 2012; Peck 1998). Burn wound infections also contribute to morbidity, lengthening recovery times, and increasing the extent of scarring (Church 2006; Oncul 2009), as well as the pain experienced by people with burns (Tengvall 2006).

All open wounds offer an ideal environment for microbial colonisation. Most wounds will contain some micro‐organisms but this will not necessarily lead to adverse events (AWMA 2011). Recently the view has developed that it is infection with sufficient or specific types of pathogenic micro‐organisms, or both, and possibly resulting biofilms (Percival 2004; Wolcott 2008) that may lead to negative outcomes and, potentially, delayed healing (Bowler 2003; Davies 2007; Madsen 1996; Trengove 1996). Biofilms are formed by bacteria that grow on a surface to form a film of cells. Growing in this way can make them more resistant to bactericidal agents. Previously it was thought that the critical factor was a threshold concentration of microbes (bioburden) (Robson 1968). However, the impact of microbial colonisation on wound healing is not independent of the host response. The ability of the host to provide adequate immune response is likely to be as critical, if not more so, in determining whether a wound becomes infected as the specifics of the flora in the wound.

People with burns have a particular vulnerability to infection, as a result of the loss of the physical barrier to infection, and the reduction in immunity mediated by the lost cells (Ninnemann 1982; Winkelstein 1984). Infections commonly occur in the acute period following the burn (Church 2006).

The spectrum of infective agents that can be present in the burn wounds varies. Nowadays, Gram‐positive bacteria such as Staphylococcus aureus (S. aureus), and Gram‐negative bacteria such as Pseudomona aeruginosa (P. aeruginosa) are the predominant pathogens (Wibbenmeyer 2006), although other micro‐organisms such as fungi, yeasts, and viruses can also be present (Church 2006; Polavarapu 2008). Multidrug‐resistant micro‐organisms, such as methicillin‐resistant S. aureus (MRSA), are frequently and increasingly identified in burns (Church 2006; DeSanti 2005; Keen 2010).

Description of the intervention

Standard care

The care for burn wounds is determined in part by their severity (depth), area, and location (National Network for Burn Care 2012). For significant injuries involving the lower layers of skin, standard care may involve a range of dressings or skin substitutes, or both, (Wasiak 2013) and more complex interventions such as hyperbaric oxygen therapy and negative pressure wound therapy (Dumville 2012; Villanueva 2004). The nature and extent of the burn wound, together with the type and amount of colonising micro‐organisms can also influence the risk of invasive infection (Bang 2002; Fitzwater 2003).

Antiseptics

Antiseptics are topical antimicrobial agents which are thought to prevent the growth of pathogenic micro‐organisms without damaging living tissue (Macpherson 2004). Applications broadly fall into two categories: lotions used for wound irrigation or cleaning, or both, with a brief contact time (unless used as a pack/soak), and products that are in prolonged contact with the wound such as creams, ointments, and impregnated dressings (BNF 2016).

Agents used primarily for wound irrigation/cleaning across wound types are commonly based on povidone‐iodine, chlorhexidine and peroxide agents. Less commonly used are traditional agents such as gentian violet and hypochlorites. Longer contact creams and ointments include fusidic acid, mupirocin, neomycin sulphate and iodine (often as cadexomer iodine). Some of these are rarely used in clinical practice. Silver‐based products such as silver sulfadiazine and silver‐impregnated dressings are increasingly used, as are honey‐based products. Aloe Vera is also sometimes used as an antiseptic although there is currently no available sterile source.

The British National Formulary (BNF) categorises antimicrobial dressings under honey‐based, iodine‐based, silver‐based, and other, which includes dressings impregnated with agents such as chlorhexidine or peroxides (BNF 2016). The choice of dressing for a burn wound is based on a number of factors including the need to accommodate movement, the minimisation of adherence to the wound surface, the prevention of infection, the ability to absorb wound fluid and maintain humidity, and the active promotion of healing (Wasiak 2013).

Antibiotics are substances that destroy or inhibit the growth of bacteria (Macpherson 2004) (normally by inhibiting deoxyribonucleic acid (DNA), protein synthesis or by disrupting the bacterial cell wall). Routine prophylaxis against infection with systemic antibiotics is not currently recommended. While it may reduce burn wound infections, or colonisation, or both, it does not decrease mortality, and may in fact increase the risk of selecting resistant micro‐organisms such as MRSA (Avni 2010; Barajas‐Nava 2013)

In contrast, antiseptics (the focus of this review) can be bactericidal (in that they kill micro‐organisms) or they can work by slowing the growth of organisms (bacteriostatic) (Macpherson 2004), but they usually work without damaging living tissue. Antiseptics can reduce the presence of other micro‐organisms such as viruses and fungi, as well as bacteria, and often work by damaging the surface of microbes (Macpherson 2004). According to the BNF (BNF 2016) antiseptics are used to reduce the presence of micro‐organisms on living tissue.

How the intervention might work

This review considers the use of antiseptics for both clinically infected and non‐infected burn wounds. The rationale for treating clinically infected wounds with antiseptic agents is to kill or slow the growth of the pathogenic micro‐organisms, thus preventing an infection from worsening and spreading (Kingsley 2004). In the case of burns, the prevention of infections, and systemic infections in particular, is especially important, as people with burns can have lowered immunity as a consequence of their injury (Church 2006). Improved healing may also result, although evidence on the association between wound healing and infection is limited (Jull 2015; O'Meara 2001; Storm‐Versloot 2010).

There is a widely held view that wounds that do not have clear signs of clinical infection, but that have characteristics such as retarded healing, may also benefit from a reduction in bacterial load (bioburden). Again, evidence for this is limited (AWMA 2011; Howell‐Jones 2005).

Why it is important to do this review

Burn wounds are a source of substantial morbidity and mortality; much of this results from the original wound becoming infected (Latenser 2007). While infections pose real risks to people with burns, the problem of antibiotic and multi‐drug resistance in bacteria continues to grow (Church 2006; DeSanti 2005; Keen 2010); alternatives to routine use of antibiotics for the minimisation of infection can be a key element of care.

There is a current published Cochrane review of antibiotics for the prevention (prophylaxis) of burn wound infection (Barajas‐Nava 2013), while a second Cochrane review of antibiotics for the treatment of infected burn wounds is now underway (Lu 2016). This review of antiseptics complements these reviews and will complete the assessment of evidence for agents with antimicrobial properties in the care of all burn wounds, whether infected or not. There will be some overlap between this review and other Cochrane and non‐Cochrane reviews of dressings for partial‐thickness burns (Wasiak 2013), and of individual agents with antiseptic properties for all types of wounds (Aziz 2012; Jull 2015; Storm‐Versloot 2010). However, this review will provide a single synthesis of the randomised evidence relating to all antiseptics for any type of burn wound.

Objectives

To assess the effects and safety of antiseptics for the treatment of burns in any care setting.

Methods

Criteria for considering studies for this review

Types of studies

We included published and unpublished randomised controlled trials (RCTs), including cluster‐RCTs, irrespective of language of report. We planned to only include crossover trials if they reported outcome data at the end of the first treatment period, prior to crossover. We excluded quasi‐randomised studies.

Types of participants

We included studies enrolling participants of any age with burn wounds. We included burns of any type, severity, extent or current infection status, managed in any care setting. We accepted authors' definitions of the category of burn represented in included trials. We included trials of participants with burns, alongside people with other types of wounds where the participants with burns constituted at least 75% of the trial population.

Types of interventions

The interventions of interest were topical antiseptic agents. We included any RCT in which the use of a specific topical antiseptic was the only systematic difference between treatment groups; where the antiseptic agent was an integral part of the dressing we allowed for this. Control regimens could have included placebo, an alternative antiseptic, another therapy such as antibiotics or isolation of the patient, standard care or no treatment. We included studies that evaluated intervention schedules, including other therapies, provided that these treatments were delivered in a standardised way across the trial arms. We excluded trials in which the presence or absence of a specific antiseptic was not the only systematic difference. We also excluded evaluations of antiseptics used to prepare for the surgical treatment of burns (i.e. where antisepsis is part of the perioperative procedure).

We anticipated that likely comparisons would include use of different antiseptic agents, in particular, the use of different types of dressings impregnated with antiseptic agents; comparisons of impregnated dressings or other antiseptic preparations with standard care; and comparison of antiseptics with topical or systemic antibiotics. We anticipated that other elements of standard care may have been co‐interventions across trial arms.

Types of outcome measures

Primary outcomes

The primary effectiveness outcome for this review was wound healing. Trialists use a range of different methods of measuring and reporting this outcome. We considered that RCTs that reported one or more of the following provided the most relevant and rigorous measures of wound healing:

• time to complete wound healing (correctly analysed using survival, time‐to‐event approaches). Ideally the outcome would be adjusted for appropriate covariates e.g. baseline wound area/degree/duration;

• proportion of wounds completely healed during follow‐up (frequency of complete healing).

We used and reported the study authors’ definitions of complete wound healing where this was available. We reported outcome measures at the latest time point available (assumed to be length of follow‐up if not specified) and the time point specified in the methods as being of primary interest (if this was different from latest time point available).

Where both the outcomes above were reported, we presented all data in a summary outcome table for reference, but focused on reporting time to healing. When time to healing was analysed as a continuous measure, but it was not clear whether all wounds healed, we documented the use of the outcome in the study, but we did not extract, summarise or use the data in any meta‐analysis.

The primary safety outcome for the review was change in wound infection status (as defined by the study authors). In the case of wounds that were considered to be clinically infected at baseline, we assessed resolution of infections. In the case of wounds that were not considered to be clinically infected at baseline, we assessed the incidence of new infections. We also assessed the incidence of septicaemia, where data permitted. We did not extract data on microbiological assays not clearly linked to a diagnosis of infection.

Secondary outcomes

We included the following secondary outcomes:

• Adverse events

  • Where reported, we extracted data on all serious adverse events and all non‐serious adverse events. We did not report individual types of adverse events other than pain (see below) or infection (see Primary outcomes).

• Health‐related quality of life

  • We included quality of life where it was reported, using a validated scale such as the SF‐36 or EQ‐5D, or a validated disease‐specific questionnaire. Ideally, reported data were adjusted for the baseline score.

• Pain (including pain at dressing change)

  • We included pain only where mean scores with a standard deviation were reported using a scale validated for the assessment of pain levels, such as a visual analogue scale (VAS).

• Resource use (when presented as a mean with standard deviation)

  • We included measures of resource use such as number of dressing changes, number of nurse visits, length of hospital stay, and need for other interventions.

• Costs associated with resource use (including estimates of cost‐effectiveness)

• Mortality (overall and infection‐related).

Search methods for identification of studies

Electronic searches

We searched the following electronic databases to identify relevant RCTs:

• the Cochrane Wounds Specialised Register (searched 26 September 2016);

• the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, 2016, Issue 8, searched 26 September 2016);

• Ovid MEDLINE (1946 to 26 September 2016);

• Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations) (searched 26 September 2016);

• Ovid Embase (1974 to 26 September 2016);

• EBSCO Cumulative Index to Nursing and Allied Health Literature (CINAHL) (1982 to 28 September 2016)

The search strategies are shown in Appendix 1.

We combined the Ovid MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision) (Lefebvre 2011). We combined the Embase search with the Ovid Embase filter developed by the UK Cochrane Centre (Lefebvre 2011). We combined the CINAHL searches with the trial filters developed by the Scottish Intercollegiate Guidelines Network (SIGN 2015). There were no restrictions with respect to language, date of publication or study setting.

We also searched the following clinical trials registries.

• ClinicalTrials.gov (www.clinicaltrials.gov/).

• WHO International Clinical Trials Registry Platform (apps.who.int/trialsearch/Default.aspx).

• EU Clinical Trials Register (www.clinicaltrialsregister.eu/).

Searching other resources

We tried to identify other potentially eligible trials or ancillary publications by searching the reference lists of retrieved included trials, as well as relevant systematic reviews, meta‐analyses and health technology assessment reports.

Data collection and analysis

Selection of studies

Two review authors independently assessed the titles and abstracts of the citations retrieved by the searches for relevance. After this initial assessment, we obtained full‐text copies of all studies considered to be potentially relevant. Two review authors independently checked the full papers for eligibility; we resolved disagreements by discussion and, where required, the input of a third review author. We obtained translation support, where necessary, for non‐English language reports. Where the eligibility of a study was unclear, we attempted to contact study authors. We recorded all reasons for exclusion of studies for which we had obtained full copies. We completed a PRISMA flowchart to summarise this process (Liberati 2009).

Where studies were reported in multiple publications/reports, we attempted to obtain all publications. Whilst we included each study only once in the review, we extracted data from all reports to ensure that we obtained all available relevant data.

Data extraction and management

We extracted and summarised details of the eligible studies. Where possible we extracted data by treatment group for the prespecified interventions and outcomes in this review. Two review authors independently extracted data; discrepancies were resolved through discussion or by consultation with a third reviewer. Where data were missing from reports, we attempted to contact the study authors and request this information.

Where we included a study with more than two intervention arms, we only extracted data from intervention and control groups that met the eligibility criteria. Where the reported baseline data related to all participants, rather than to those in relevant treatment arms, we extracted the data for the whole trial and noted this. We collected outcome data for relevant time points as described in the Types of outcome measures.

Where possible, we extracted the following data:

• bibliographic data, including date of completion/publication;

• country of origin;

• unit of randomisation (participant/wound);

• unit of analysis;

• trial design e.g. parallel, cluster;

• care setting;

• number of participants randomised to each trial arm and number included in final analysis;

• eligibility criteria and key baseline participant data including cause, depth, extent (area/proportion of total body surface area (TBSA)) and location of burns; ages of participants, and whether they had a diagnosis of infection at baseline;

• details of treatment regimen received by each group;

• duration of treatment;

• details of any co‐interventions;

• primary and secondary outcome(s) (with definitions and, where applicable, time points);

• outcome data for primary and secondary outcomes (by group);

• duration of follow‐up;

• number of withdrawals (by group) and number of withdrawals (by group) due to adverse events;

• publication status of study;

• source of funding for trial.

Assessment of risk of bias in included studies

Two review authors independently assessed included studies using the Cochrane tool for assessing risk of bias (Higgins 2011a). This tool addresses specific domains: sequence generation, allocation concealment, blinding, incomplete data, selective outcome reporting and other issues. In this review we recorded issues with unit of analysis, for example where a cluster trial has been undertaken but analysed at the individual level in the study report.

We assessed blinding of outcome assessment and completeness of outcome data for each of the review outcomes separately. We presented our assessment of risk of bias using two 'Risk of bias' summary figures; one is a summary of bias for each item across all studies, and a second shows a cross‐tabulation of each trial by all of the risk of bias items.

We summarised a study’s risk of selection bias, detection bias, attrition bias, reporting bias and other bias. In many of the comparisons included in this review, we anticipated that blinding of participants and personnel may not be possible. For this reason the assessment of the risk of detection bias focused on whether blinded outcome assessment was reported (because wound healing can be a subjective outcome, it can be at high risk of measurement bias when outcome assessment is not blinded). For trials using cluster randomisation, we also planned to consider risk of bias for recruitment bias, baseline imbalance, loss of clusters, incorrect analysis and comparability with individually‐randomised trials (Higgins 2011b) (Appendix 3).

Measures of treatment effect

We reported time‐to‐event data (e.g. time‐to‐complete wound healing) as hazard ratios (HRs) when possible, in accordance with the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). If studies reporting time‐to‐event data (e.g. time to healing) did not report an HR, then, when feasible, we estimated this using other reported outcomes, such as numbers of events, through the application of available statistical methods (Parmar 1998; Tierney 2007). This included deriving an HR from data reported for multiple time points, where at least three time points were reported. Where no HR could be calculated, we reported dichotomous data at the latest time point. For dichotomous outcomes, we calculated the risk ratio (RR) with 95% confidence intervals (CIs). For continuous outcome data, we used the difference in means (MD) with 95% CIs for trials that used the same assessment scale. When trials used different assessment scales, we used the standardised difference in means (SMD) with 95% CIs.

Unit of analysis issues

Where studies were randomised at the participant level and outcomes measured at the wound level, for example for wound healing, we treated the participant as the unit of analysis when the number of wounds assessed appeared to be equal to the number of participants (e.g. one wound per person).

One unit of analysis issue that we anticipated was that randomisation may have been carried out at the participant level, with the allocated treatment used on multiple wounds per participant (or perhaps only on some participants), but data were presented and analysed per wound (clustered data).

In cases where included studies contained some or all clustered data, we reported this, noting whether data had been (incorrectly) treated as independent. We recorded this as part of the 'Risk of bias' assessment.

We also included studies with the split‐body design where either people with two similar burn wounds were enrolled and each burn wound was randomised to one of the interventions, or where one half of a wound was randomised to one treatment and the other half to a different treatment. These approaches are similar to the 'split‐mouth' approach (Lesaffre 2009). These studies should be analysed using paired data which reflects the reduced variation in evaluating different treatments on the same person. However, it was often not clear whether such analysis had been undertaken. This lack of clarity is noted in the 'Risk of bias' assessment and in the notes in the Characteristics of included studies table

We adopted a pragmatic but conservative post‐hoc approach to analyses including clustered and paired data. We included such studies in meta‐analyses where possible (where unadjusted clustered data would produce too‐narrow CIs and unadjusted paired data too‐wide CIs). We undertook a post‐hoc sensitivity analysis to explore the impact of including data that had been inappropriately unadjusted. Where the sensitivity analysis produced a materially different result to the primary analysis, we used this as the basis for the GRADE assessment and the 'Summary of findings' table. Where we pooled studies with paired data with one other trial, we also reported the results of both trials individually, and where a paired data study was the sole trial reporting outcome data, we noted the issues related to its design. We also noted where these trials were included in meta‐analyses but did not contribute weight to the analyses due to zero events or lack of measures of variance.

Dealing with missing data

It is common to have data missing from trial reports. Excluding participants from the analysis post randomisation, or ignoring participants who are lost to follow‐up compromises the randomisation and potentially introduces bias into the trial. If it was thought that study authors might be able to provide some missing data, we attempted to contact them; however, data are often missing because of loss to follow‐up. In individual studies, when data on the proportion of burns healed were presented, we assumed that randomly‐assigned participants not included in an analysis had an unhealed wound at the end of the follow‐up period (i.e. they were considered in the denominator but not in the numerator). When a trial did not specify participant group numbers before dropout, we presented only complete case data. For time‐to‐healing analysis using survival analysis methods, dropouts should be accounted for as censored data. Hence all participants will be contributing to the analysis. We acknowledge that such analysis assumes that dropouts are missing at random and there is no pattern of missingness. We presented data for all secondary outcomes as a complete case analysis.

For continuous variables (e.g. length of hospital stay) and for all secondary outcomes, we presented available data from the study reports/study authors and did not impute missing data. Where measures of variance were missing, we calculated these, wherever possible (Higgins 2011a). If calculation was not possible, we contacted the study authors. Where these measures of variation remained unavailable and we could not calculate them, we excluded the study from any relevant meta‐analyses that we conducted.

Assessment of heterogeneity

Assessment of heterogeneity can be a complex, multi‐faceted process. Firstly, we considered clinical and methodological heterogeneity; that is the degree to which the included studies varied in terms of participants, interventions, outcomes, and characteristics such as length of follow‐up. We supplemented this assessment of clinical and methodological heterogeneity by information regarding statistical heterogeneity ‐ assessed using the Chi² test (we considered a significance level of P < 0.10 to indicate statistically significant heterogeneity) in conjunction with the I² statistic (Higgins 2003). I² examines the percentage of total variation across RCTs that is due to heterogeneity rather than chance (Higgins 2003). Very broadly, we considered that I² values of 25%, or less, may mean a low level of heterogeneity (Higgins 2003), and values of 75% or more, indicated very high heterogeneity (Deeks 2011). Where there was evidence of high heterogeneity, we attempted to explore this further (see Data synthesis).

Assessment of reporting biases

Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results. Publication bias is one of a number of possible causes of 'small study effects', that is, a tendency for estimates of the intervention effect to be more beneficial in smaller RCTs. Funnel plots allow a visual assessment of whether small study effects may be present in a meta‐analysis. A funnel plot is a simple scatter plot of the intervention effect estimates from individual RCTs against some measure of each trial’s size or precision (Sterne 2011). Funnel plots are only informative when there are a substantial number of studies included in an analysis; we had planned to present funnel plots for meta‐analyses that included at least 10 RCTs using Review Manager 5 (RevMan 5) (RevMan 2014) but there were no analyses with sufficient studies.

Data synthesis

We combined details of included studies in narrative review according to the comparison between intervention and comparator, the population and the time point of the outcome measurement. We considered clinical and methodological heterogeneity and undertook pooling when studies appeared appropriately similar in terms of burn type and severity, intervention type and antibacterial agent, duration of treatment and outcome assessment.

In terms of a meta‐analytical approach, in the presence of clinical heterogeneity (review author judgement), or evidence of statistical heterogeneity, or both, we used a random‐effects model. We planned to only use a fixed‐effect approach when clinical heterogeneity was thought to be minimal and statistical heterogeneity was estimated as non‐statistically significant for the Chi2 value and 0% for the I2 assessment (Kontopantelis 2013). We adopted this approach as it is recognised that statistical assessments can miss potentially important between‐study heterogeneity in small samples, hence the preference for the more conservative random‐effects model (Kontopantelis 2012). Where clinical heterogeneity was thought to be acceptable, or of interest, we considered conducting meta‐analysis even when statistical heterogeneity was high, but attempted to interpret the causes behind this heterogeneity and considered using meta‐regression for that purpose, if possible (Thompson 1999; Thompson 2002).

We presented data using forest plots, where possible. For dichotomous outcomes we presented the summary estimate as a RR with 95% CIs. Where continuous outcomes were measured in the same way across studies, we planned to present a pooled MD with 95% CIs; we pooled SMD estimates where studies measured the same outcome using different methods. For time‐to‐event data, we plotted (and, where appropriate, pooled) estimates of HRs and 95% CIs, as presented in the study reports, using the generic inverse variance method in RevMan 5 (RevMan 2014). Where time to healing was analysed as a continuous measure, but it was not clear if all wounds healed, we documented use of the outcome in the study, but did not summarise the data or use the data in any meta‐analysis.

'Summary of findings' tables

We presented the main results of the review in 'Summary of findings' tables. These tables present key information concerning the quality of the evidence, the magnitude of the effects of the interventions examined and the sum of available data for the main outcomes (Schünemann 2011a). The 'Summary of findings' tables also include an overall grading of the evidence related to each of the main outcomes using the GRADE approach. The GRADE approach defines the 'certainty' of a body of evidence as the extent to which one can be confident that an estimate of effect or association is close to the true quantity of specific interest. The certainty of a body of evidence involves consideration of within‐trial risk of bias (methodological quality), directness of evidence, heterogeneity, precision of effect estimates and risk of publication bias (Schünemann 2011b). We presented the following outcomes in the 'Summary of findings' tables:

• time‐to‐complete wound healing, when analysed using appropriate survival analysis methods;

• proportion of wounds completely healing during the trial period;

• mean time to healing when all wounds healed;

• changes in clinical infection status;

• adverse events.

Where comparisons had limited available data for specified outcomes we did not generate a 'Summary of findings' table for this comparison. Instead we decided to present these data together with GRADE judgements in an additional table, in order to keep the 'Summary of findings' tables section of the review manageable and improve readability.

In terms of the GRADE assessment, when making decisions for the risk of bias domain we downgraded only when studies had been classed at high risk of bias for one or more domains. We did not downgrade for unclear risk of bias assessments. In assessing the precision of effect estimates we assessed the size of confidence intervals, downgrading twice for imprecision when there were very few events and CIs around effects included both appreciable benefit and appreciable harm. We considered CI to be especially fragile where there were fewer than 50 participants; event rates were also considered in determining fragility.

Subgroup analysis and investigation of heterogeneity

When possible, we planned to perform subgroup analyses to explore the effect of interventions in children under the age of 18, in adults, and in older adults (aged over 65 years). When possible, we also planned to use subgroup analyses to assess the influence of burn size and depth on effect size. If there had been sufficient data these analyses would have assessed whether there were differences in effect sizes for burns of different depths.

When possible, we planned to perform subgroup analyses to explore the influence of risk of bias on effect size. We planned to assess the influence of removing from meta‐analyses studies classed as having high and unclear risk of bias. These analyses would have only included studies that were assessed as having low risk of bias in all key domains, namely, adequate generation of the randomisation sequence, adequate allocation concealment, and blinding of outcome assessor for the estimates of treatment effect.

Elements of this Methods section are based on the standard Cochrane Wounds protocol template.

Results

Description of studies

Results of the search

The search identified a total of 1565 records after duplicates were removed, of which we assessed 214 records as full texts (Figure 1).

Included studies

We included 56 studies reported in 66 publications, with a total of 5807 randomised participants. Most studies had two intervention groups but two studies had three arms, each study evaluating two relevant comparisons (Chen 2006; Thomas 1995), one study had four arms and evaluated five relevant comparisons (Li 1994) and one (Piccolo‐Daher 1990) had five arms and evaluated two relevant comparisons. A number of the studies enrolled participants with two comparable burn wounds and randomly assigned the wounds to the interventions (that is randomisation was at the wound rather than participant level).

Included studies assessed the following types of comparisons.

• Comparisons between antiseptics and topical antibiotics

• Comparisons between two antiseptics

• Comparisons between antiseptics and treatments without antimicrobial properties.

The main groups of interventions and the direct comparisons between them are shown in Figure 2 and listed in Table 9.

Network of included treatment types

1

Summary of comparisons

Risk of bias in included studies

No studies had a low risk of bias for all domains. We judged only one study (Tang 2015) to be at low risk of bias across all except one domain, where there was an unclear risk of bias. All other studies had an unclear or high risk of bias for two or more domains. There were 17 studies with one domain classed at high risk of bias and we rated three of these studies as being at high risk of bias in more than one domain. Most studies had multiple domains which were at an unclear risk of bias. For only two domains (attrition bias and reporting bias) did we consider a majority of the studies to be at low risk of bias. Figure 3 and Figure 4 illustrate the predominance of unclear judgements across the domains.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

Allocation

Generation of randomisation sequence and concealment of allocation were not well reported. Most studies had an unclear risk of bias with allocation concealment, especially, poorly documented.

Blinding

Blinding of outcome assessment was largely unclear although several studies were clearly not blinded for their primary outcomes. We judged a smaller number to be at low risk of bias.

Incomplete outcome data

Most studies were at low risk of attrition bias with all participants accounted for or only a small number missing from analyses. Approximately 10% of studies were at high risk of bias and 20% were unclear.

Selective reporting

Just over half the studies were at low risk of bias, we judged three to be at high risk of bias and the remainder were unclear.

Other potential sources of bias

A minority of studies were sufficiently well‐reported and conducted for us to be confident that they were at low risk of other sources of bias. While we judged only two studies to be at high risk of bias due to unit of analysis issues (Huang 2007; Thomas 1995), over half the studies were too poorly reported for us to be clear that there were no other potential sources of bias. None of the ten studies which used intra‐individual designs for both randomisation and analysis made it clear whether they had used appropriate analytical methods for the paired data. We judged these studies to be unclear for this domain in the 'Risk of bias' assessment (Homann 2007; Khorasani 2009; Liao 2006; Malik 2010; Nasiri 2016; Piatkowski 2011; Radu 2011; Varas 2005; Yang 2013; Zhou 2011). The effect of failure to account for pairing would be to produce wider confidence intervals than the appropriate analysis. Zhou 2011 may be at particularly high risk of carry‐over effects from one intervention to another as it randomised burn areas rather than discrete burns.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4; Table 5; Table 6; Table 7; Table 8

Summary of findings for the main comparison

Silver‐based antiseptics versus topical antibiotics

Summary of findings 2

Honey versus topical antibiotics

Summary of findings 3

Aloe vera versus topical antibiotics

Summary of findings 4

Iodine versus topical antibiotics

Summary of findings 5

Silver versus non‐antibacterial

Summary of findings 6

Honey versus non‐antibacterial

Summary of findings 7

Chlorhexidine versus non‐antibacterial

Summary of findings 8

Iodine versus non‐antibacterial

Individual study outcome data are shown in Table 10 (wound healing); Table 11 (wound infection); and Table 12 (secondary outcomes).

2

Summary of data for wound healing

3

Summary of reported data for infection

4

Summary of secondary outcome data for comparisons

Comparison between antiseptics and topical antibiotics

1. Silver‐based antiseptic treatments versus topical antibiotics (16 studies, 1368 participants)

Silver‐based treatments included silver foam dressings, silver hydrogel dressings, silver alginate dressing, nanocrystalline silver dressing, silver hydrofibre dressing and silver nitrate. In each case the topical antibiotic used was silver sulfadiazine (SSD). Four studies randomised a total of 373 participants and assessed nanocrystalline silver (Chen 2006; Huang 2007; Muangman 2006; Varas 2005). Four studies with a total of 318 participants assessed silver foam dressings (Silverstein 2011; Tang 2015; Yarboro 2013; Zhou 2011). Silver hydrogel dressings were assessed in three studies (Adhya 2015; Glat 2009; Gong 2009; 191 participants) as were silver hydrofibre dressings (Abedini 2013; Caruso 2006; Muangman 2010; 201 participants). Single studies assessed a silver alginate dressing (Opasanon 2010, 65 participants) and silver nitrate treatment (Liao 2006, 120 participants).

Two studies included only children (Glat 2009; Zhou 2011) and three included only adults (Gong 2009, Huang 2007; Varas 2005). The remaining studies included both adults and children or did not report this. Most studies included recent burns described as second‐degree or partial‐thickness but one (Huang 2007) included only residual burns, unhealed despite previous treatment. The percentage of total body surface area affected (TBSA) was below 40% in all except one study (Adhya 2015 included burns up to 60% TBSA) and several studies imposed lower limits of 10%, 15% or 25%.

Four studies used burns rather than participants as the unit of analysis (Huang 2007; Liao 2006; Varas 2005; Zhou 2011). In Huang 2007 the randomisation was at the level of the participant but the analysis was conducted at the level of the burn wound, that is multiple burns on the same participants were treated with the same treatment and outcome data for the different wounds analysed (clustered data); the other studies employed an intra‐individual (split‐body) design for both randomisation and analysis; in each case it was not clear whether the analysis had adjusted appropriately. Although Tang 2015 enrolled participants with multiple burns, a single burn was selected at study enrolment and both randomisation and analysis were at the participant level.

Primary outcome: wound healing

Table 1

Most studies reported some data on wound healing with this being presented in different ways. There were three studies (259 participants) with sufficient data to calculate a HR for healing and we pooled these data (Caruso 2006; Glat 2009; Tang 2015). In this analysis, on average, the use of silver‐containing antiseptics treatment (mainly dressings) showed no clear difference in time to healing compared with SSD; the estimate is imprecise, with CIs spanning benefits and harms (HR 1.25, 95% CI 0.94 to 1.67; I2 = 0%) Analysis 1.1 (low certainty evidence, downgraded twice for imprecision; confidence intervals included both the possibility of a 6% decrease and a 67% increase in the 'chance' of healing).

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 1 Wound healing (hazard ratio).

Mean time to wound healing was reported in ten studies where it seemed that all wounds had healed. On average, silver‐containing antiseptic treatments (mainly dressings) may decrease slightly the mean time to healing of burns compared with SSD (MD ‐3.33 days; 95% CI ‐4.96 to ‐1.70; I2 = 87%) (low certainty evidence downgraded once for risks of bias (variously detection, selection, reporting and other sources of bias across four of the studies and 30% of the analysis weight) and once for inconsistency due to high heterogeneity). Although statistical heterogeneity was high, all studies had the same direction of effect and favoured silver‐containing antiseptics Analysis 1.2. This was based on nine studies; Silverstein 2011 did not report measures of variance. We used a post‐hoc sensitivity analysis to explore the impact of including studies which may not have adjusted for clustered or intra‐individual designs. Excluding Huang 2007; Liao 2006; and Zhou 2011 from the analysis resulted in a lower level of heterogeneity (I2 = 36%) and a slightly larger estimate of effect (MD ‐4.53 days, 95% CI ‐5.74 to ‐3.32); excluding only trials with intra‐individual designs (Liao 2006; and Zhou 2011) or only the trial with unit of analysis issues (Huang 2007) also produced little difference.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 2 Wound healing (mean time to healing).

The RR for short‐term follow‐up (maximum 28 days) suggested that on average the use of silver‐containing antiseptics may lead to a small difference in number of healing events over one month compared with SSD: RR 1.17 (95% CI 1.00 to 1.37; I2 = 45%) (Analysis 1.3) (low certainty evidence, downgraded once due to risk of detection bias in two studies and selection bias in one study; and once due to imprecision).

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 3 Wound healing (risk ratio) up to 28 days.

Primary outcome: infection

Table 1

Incident infections were reported in three studies: Caruso 2006; Glat 2009; Muangman 2006. Tang 2015 reported new signs of wound inflammation which we grouped with the incident infections. It is uncertain whether use of silver‐containing antiseptics prevents infection compared with SSD because the certainty of the evidence is very low: RR 0.84 (95% 0.48 to 1.49; I2 = 0%) Analysis 1.4 (very low certainty evidence, downgraded once each for risk of bias (variously reporting, detection and selection), imprecision and indirectness). Huang 2007 reported bacterial clearance rates (including for specific strains including MRSA); these data are noted in Table 11.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 4 Infection (up to 4 weeks or NR).

Secondary outcome: adverse events

Table 1

Eight studies reported some data on adverse events (Caruso 2006; Glat 2009; Gong 2009; Huang 2007; Silverstein 2011; Tang 2015; Varas 2005; Zhou 2011); six studies reported the proportion of all participants with adverse events.

There was no clear difference in the incidence of adverse events between silver‐containing antiseptics and SSD in the number of participants with adverse events; the estimate is imprecise with wide CIs spanning benefits and harms: RR 0.86 (95% CI 0.63 to 1.18; I2 = 0%) (Analysis 1.5) (low certainty evidence, downgraded once for risks of bias (variously detection, selection, reporting and other sources of bias) across five of the trials, and once for imprecision). The analyses included six trials, three of which reported that there were no events. We considered a post‐hoc sensitivity analysis to explore the impact of including studies that may not have adjusted for clustered or intra‐individual designs, however, both Huang 2007 and Zhou 2011 reported zero events and therefore did not contribute weight to the analysis.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 5 Adverse events (14‐28 days).

Other trials reported data relating to withdrawals or specific event types (Gong 2009) including serious adverse events and withdrawals due to adverse events: these data are not included in the main analysis but are reported separately (Analysis 1.6; Table 12). Because one of the two trials (Silverstein 2011; Varas 2005) reporting withdrawals due to adverse events had an intra‐individual design we both present pooled data for this analysis and report the results of the trials separately (Table 12).

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 6 Withdrawals due to adverse events (21 days or NR).

Secondary outcome: pain

Eleven trials reported some data on pain. The most commonly reported measures were pain in general (or at an unspecified time) and pain at dressing change. Caruso 2006; Glat 2009; Muangman 2010;Tang 2015; and Yarboro 2013 reported usable data on pain at dressing change. Gong 2009 reported only the presence of pain at dressing change in the SSD group and its absence in the silver‐based antiseptic group, and Silverstein 2011 only that there was no significant difference between the groups.

Silver‐based antiseptic treatments may on average slightly reduce pain at dressing change compared with SSD, SMD ‐1.20 (95% CI ‐1.92 to ‐0.49; I2 = 81%) (low certainty evidence, downgraded once for imprecision and once for inconsistency). There was significant statistical heterogeneity between the studies, but all of the trials reported lower pain levels in the silver antiseptic group than in the SSD group Analysis 1.7.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 7 Pain at dressing change (up to 28 days or NR).

A general measure of pain was reported by three trials (Muangman 2006; Opasanon 2010; Varas 2005). Silver‐based antiseptic treatments may, on average, slightly reduce generally reported pain compared with SSD. Pain scores may on average be slightly lower in participants treated with silver dressings; the SMD was ‐1.66 (95% CI ‐2.06 to ‐1.27; I2 = 0%). Analysis 1.8 (low certainty evidence, downgraded once for risk of reporting bias or attrition bias affecting over half the participants and once for imprecision). We used a post‐hoc sensitivity analysis to explore the impact of including studies that may not have adjusted for clustered or intra‐individual designs. Excluding Varas 2005 resulted in no change to the estimate of effect but slightly wider confidence intervals (SMD ‐1.66, 95% CI ‐2.74 to ‐0.64; I2 = 0%). Further pain‐related measures, which could not be analysed here, are reported in Table 12.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 8 Pain (time/follow‐up not specified).

Secondary outcome: mortality

Three trials reported mortality: Caruso 2006; Silverstein 2011 and Muangman 2006. It is uncertain whether silver‐containing antiseptic treatments have an effect on mortality. The RR was 1.59 (95% CI 0.20 to 12.64; I2 = 0%) Analysis 1.9. (very low certainty evidence, downgraded once for risks of detection bias and reporting bias and twice for imprecision; two trials at risk of detection bias, one at risk of reporting bias).

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 9 Mortality (21 days or NR).

Secondary outcome: resource use

Number of dressing changes was reported by six trials (Caruso 2006; Glat 2009; Opasanon 2010; Silverstein 2011; Tang 2015; Yarboro 2013). Participants treated with silver‐based antiseptics (dressings) may require fewer dressing changes compared with those treated with SSD. Data from four studies (Silverstein 2011 and Tang 2015 did not report measures of variance) suggests that, on average, silver‐containing antiseptics (dressings) may reduce the number of dressing changes, MD ‐7.56 dressing changes (95% CI ‐12.09 to ‐3.04; I2 = 84%) Analysis 1.10 (low certainty evidence, downgraded once for risks of detection and selection bias affecting three trials with 45% of analysis weight, and once for imprecision). The number of minutes of nursing time required was also reported by Opasanon 2010, this also showed that there may be a small benefit to silver‐based antiseptics (difference in means ‐4.82 minutes, 95% CI ‐7.42 to ‐2.22) (low certainty evidence, downgraded twice for imprecision) (Table 12). Silverstein 2011 reported mean time to discharge but without measures of variance; the data are shown in Table 12 but are not further analysed.

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 10 Resource use (number of dressings) (up to 28 days or NR).

Secondary outcome: costs

Four trials reported data for total costs of treatment (Abedini 2013; Caruso 2006; Muangman 2010; Silverstein 2011 (based on a subset of 20 participants' data)). It is uncertain whether or not silver‐based antiseptic dressings are cheaper overall than SSD. The pooled difference in means across the four trials was USD ‐117.18 (95% CI ‐280.02 to 45.67; I2 = 68%) Analysis 1.11. This is very low certainty evidence downgraded once for risk of detection bias in three of the four studies (accounting for over 50% of participants) and twice for imprecision (confidence intervals included both possibilities of cost reduction (USD 280) and increase (USD 46)). Cost‐effectiveness data from Caruso 2006 and Silverstein 2011 also showed uncertainty as to whether silver dressings were more cost‐effective than SSD (very low certainty evidence, downgraded twice for imprecision and once for risk of detection bias in both studies) (Table 12).

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 11 Costs (21 days or NR).

Summary of comparison

Low certainty evidence reporting the hazard or 'chance' of healing over time suggested that there may be a small benefit for burns treated with silver‐based antiseptics (mainly silver‐containing dressings) compared with SSD but confidence intervals were wide, spanning both benefits and harms so clear differences between treatments are not apparent. Low certainty evidence also showed that mean time to healing may be somewhat (3 days) shorter with silver‐based antiseptics compared to SSD. There is very low certainty evidence on infection incidence and mortality, meaning that it is unclear what the effect of the different interventions may be. There is low certainty evidence on adverse events suggesting that there may be little or no difference between the treatments. Pain scores may be slightly lower in participants treated with silver compared with SSD (low certainty evidence). Table 1

2. Honey or honey‐based dressings versus topical antibiotic (11 studies, 856 participants)

Nine studies used honey (variously described as pure, undiluted, unprocessed) (Baghel 2009; Malik 2010; Mashhood 2006; Memon 2005; Maghsoudi 2011; Sami 2011; Subrahmanyam 1991; Subrahmanyam 1998; Subrahmanyam 2001), one a honey dressing (Bangroo 2005) and one olea which contains honey and olive and sesame oils (Zahmatkesh 2015). Eight studies used SSD as the comparator and two used mafenide acetate (Maghsoudi 2011; Zahmatkesh 2015).

Nine studies included a mix of adults and children (Malik 2010; Mashhood 2006; Memon 2005; Maghsoudi 2011; Sami 2011; Subrahmanyam 1991; Subrahmanyam 1998; Subrahmanyam 2001; Zahmatkesh 2015), one included only adults (Baghel 2009), and one only children (Bangroo 2005). Six studies included participants with burns less than 40% TBSA (Malik 2010; Memon 2005; Maghsoudi 2011; Subrahmanyam 1991; Subrahmanyam 1998; Subrahmanyam 2001); two studies specified less than 50% TBSA (Baghel 2009; Bangroo 2005) and one less than 15% TBSA (Mashhood 2006). One study (Malik 2010) used an intra‐individual design and randomised burns on each participant to the treatments.

Primary outcome: wound healing

Table 2

All studies reported some measure of wound healing. One study reported the mean time to healing of all wounds but with no measure of variance reported (Baghel 2009). A second study reported full data on only one intervention group (Bangroo 2005). Zahmatkesh 2015 reported the median time to formation of granulation tissue. These data are all presented in Table 10 but are not analysed further.

We could calculate HRs for healing for five studies (Maghsoudi 2011; Malik 2010; Mashhood 2006; Memon 2005; Sami 2011). Honey probably on average reduces time to healing compared with topical antibiotics: HR 2.45 (95% CI 1.71 to 3.52; I2 = 66%) Analysis 2.1 (moderate certainty evidence, downgraded once due to imprecision). This would correspond to an additional 278 (95% CI 185 to 332) more burns healed over time for every 1000 burns treated. We used a post‐hoc sensitivity analysis to explore the impact of including the study with an intra‐individual design (Malik 2010). The results of this sensitivity analysis differed little from the main analysis (HR 2.31, 95% CI 1.43 to 3.71; I2 = 67%).

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 1 Wound healing (hazard ratio).

Six studies reported the mean time to healing where all wounds healed but for two studies (Baghel 2009; Memon 2005) no measure of variance was available. Honey may slightly shorten the mean number of days to wound healing compared with topical antibiotics). Based on analysis of four studies, the average mean time to healing was ‐3.79 days (95% CI ‐7.15 to ‐0.43; I2 = 96%) shorter in participants treated with honey compared with those treated with SSD, and all studies showed the same direction of effect despite high statistical heterogeneity Analysis 2.3. We used a post‐hoc sensitivity analysis to explore the impact of including the study with an intra‐individual design (Malik 2010). The estimate of effect was increased slightly, but wider confidence intervals included the possibility of a small increase in mean time to healing as well as a decrease (MD ‐4.36 days, 95% CI ‐8.90 to 0.17; I2 = 95%). This would be very low certainty evidence, downgraded twice for imprecision and once for inconsistency; in order to be conservative we have adopted the GRADE assessment based on the sensitivity analysis because it ascribes less certainty to the findings than that based on the main analysis.

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 3 Wound healing (mean time to healing).

The RR for short‐term follow up (maximum 21 days) also suggested that, on average honey, probably leads to more short‐term healing events than topical antibiotic treatment: RR 2.18 (95% CI 1.15 to 4.13; I2 = 94%). Over a longer period of up to 60 days the RR was 1.65 (95% CI 0.99 to 2.76; I2 =99%), including the data from the last time points of Mashhood 2006 and Sami 2011. Data from a study which used different time points for the two groups were not included but contribute to the HR (Memon 2005). In each case this is low certainty evidence, downgraded once for imprecision and once for inconsistency.

Primary outcome: infection

Table 2

Change in infection status

Eight studies comparing honey with topical antibiotics reported some measure of change in infection status. Four reported incident infection (Malik 2010; Maghsoudi 2011; Subrahmanyam 1998; Zahmatkesh 2015); three reported persistent infection (Sami 2011; Subrahmanyam 1991; Subrahmanyam 2001) and one reported time for swab cultures to become negative (Mashhood 2006) but with no measures of variance; these data are reported in Table 11 but are not further analysed. Most studies used a measure of infection based on swab cultures which is not a measure of clinical infection. Only Maghsoudi 2011, which compared honey with mafenide acetate, reported incidence of new clinical signs of infections (at 7 and 21 days).

Incident infections

It is uncertain if fewer burns treated with honey may become infected compared with those treated with topical antibiotics (SSD or mafenide acetate) when assessed at time points between seven and 24 days. The RR was 0.16 (95% 0.08 to 0.34; I2 = 0%) Analysis 2.4. This is very low certainty evidence, downgraded for twice for indirectness in all studies except Maghsoudi 2011 and once for imprecision due to low numbers of events. It was unclear if the analysis in Malik 2010 was adjusted for paired data. Excluding this study in a post‐hoc sensitivity analysis did not materially change the result; the RR was 0.09 (95% 0.02 to 0.35; I2 = 0%).

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 4 Incident infection (up to 24 days).

Persistent infections

It is uncertain if wounds may be more likely to become infection free at 15 (Subrahmanyam 1991) or 21 days (Subrahmanyam 2001) in groups treated with honey compared with those treated with SSD. The RR was 0.10 (95% CI 0.05, 0.19; I2 = 0%) Analysis 2.5 (very low certainty evidence, downgraded once for imprecision and twice for indirectness in all studies except Maghsoudi 2011). Sami 2011 reported the proportion of participants with continuing positive cultures at multiple time points up to six weeks, at which point all were culture negative (Table 11).

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 5 Persistent positive swabs (up to 21 days).

Secondary outcome: adverse events

Table 2

Three studies comparing honey with topical antibiotics reported adverse events for all participants (Maghsoudi 2011; Mashhood 2006; Subrahmanyam 2001). Other trials reported only individual types of events but it was very unclear whether these data related to the number of participants experiencing adverse events or whether multiple events may have been reported for some individuals. These data are noted in Table 12 but are not analysed further. It is uncertain whether fewer participants treated with honey experience adverse events compared with those treated with SSD. There were no events in two trials and two events in the topical antibiotics group in the other trial; the RR was 0.20 (95% CI 0.01 to 3.97; I2 not calculable) Analysis 2.6. This is very low certainty evidence, downgraded once because of risks of detection bias in Mashhood 2006 and twice because of imprecision.

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 6 Adverse events (time points between 21 days and 6 weeks).

Secondary outcome: resource use

One study (Subrahmanyam 2001) reported on the length of hospital stay in participants treated with honey or SSD. There is probably a shorter length of stay in participants treated with honey compared with SSD (difference in means ‐10.30 days, 95% CI ‐10.95 to ‐9.65) (Table 12). This is moderate certainty evidence, downgraded once for imprecision. A second study (Sami 2011) reported the amounts of honey or SSD required per dressing per percentage area burned. No measures of variance were reported; these data are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Secondary outcome: costs

Two studies (Mashhood 2006; Sami 2011) reported on costs of treating burns with honey or SSD but did not report any measure of variance; these data are reported in Table 12 but are not further analysed; without an estimate of effect it is difficult to provide a GRADE assessment for the outcome.

Summary of comparison

Honey on average probably reduces the time to healing compared with topical antibiotics, assessed by evidence reporting the hazard or 'chance' of healing over time (moderate certainty evidence). The mean time to healing may, on average, be reduced in wounds treated with honey compared with topical antibiotics (low certainty evidence). Compared with topical antibiotics, honey may, on average, increase the number of healing events assessed over the short term (up to 3 weeks) but it is unclear whether this is still the case when studies with longer follow‐up are included (low certainty evidence). It is unclear if there are fewer infections in wounds treated with honey compared with topical antibiotics, and whether fewer initial infections persist (very low certainty evidence). It is uncertain whether the incidence of adverse events differs between groups (very low certainty evidence). Table 2

3. Aloe Vera versus topical antibiotics (5 studies, 338 participants)

Four studies compared Aloe Vera to SSD (Khorasani 2009; Panahi 2012; Shahzad 2013; Thamlikitkul 1991) and one compared it to framycetin cream (Akhtar 1996). The Aloe Vera was administered in a variety of creams, gel or dressings. The concentration of Aloe Vera was 0.5% in the one study that reported this (Khorasani 2009).

Three studies included mostly adults (Khorasani 2009; Panahi 2012; Shahzad 2013), one did not report participant age (Akhtar 1996) and in one the mean ages suggested a mix of adults and children (Thamlikitkul 1991). Inclusion criteria for TBSA of burns ranged from less than 5% (Panahi 2012) to less than 40% (Akhtar 1996). One study (Khorasani 2009) used an intra‐individual design.

Primary outcome: wound healing

Table 3

Four studies reported data on wound healing. Three reported mean time to healing of all wounds (Akhtar 1996; Khorasani 2009; Shahzad 2013). Akhtar 1996 did not report a measure of variance, so these data are reported in Table 10 but are not further analysed. Based on the pooled data from the remaining two studies it is uncertain whether there is a difference in mean time to healing between treatments: average difference in means was ‐7.79 days (95% CI ‐17.96 to 2.38; I2 = 94%) Analysis 3.1. This is very low certainty evidence, downgraded once for risk of detection bias in a study accounting for 48% of the weight in the analysis (Shahzad 2013), once for inconsistency and twice for very high levels of imprecision. The confidence intervals included the possibility of healing time being shorter by almost 18 days or being two days longer. Because Khorasani 2009 used an intra‐individual design, we also note the separate MD for this study (MD ‐2.85 days, 95% CI ‐4.04 to ‐1.66) and Shahzad 2013 (MD ‐13.24 days, 95% CI ‐17.91 to ‐8.57).

Analysis

Comparison 3 Aloe vera vs topical antibiotics, Outcome 1 Wound healing (mean time to healing).

One study reported the proportion of wounds healed at 26 days (Thamlikitkul 1991). It is unclear whether Aloe Vera may alter the number of healing events compared with SSD; confidence intervals were wide spanning both benefits and harms so clear differences between treatments are not apparent (RR 1.41, 95% CI 0.70 to 2.85) (Table 10). This is low certainty evidence, downgraded twice for serious imprecision; confidence intervals included the possibility of both a 30% reduction and a 285% increase in the chance of wound healing. None of the studies reported sufficient information for us to calculate an HR for wound healing.

Primary outcome: infection

Table 3

Three studies reported data on the incidence of infections at different time points (Khorasani 2009 (24 days); Panahi 2012 (14 days); Shahzad 2013 (unclear time point)).

It is uncertain whether there is a difference between the groups (RR 0.93, 95% CI 0.26 to 3.34; I2 = 0%) Analysis 3.2 . This is very low certainty evidence, downgraded once for risk of detection bias because 84% of the analysis weight was represented by Shahzad 2013, which had a high risk of detection bias, and twice for imprecision. Very wide confidence intervals included both the possibility of lower (by 74%) or much higher (by over 300%) infection rates in the Aloe Vera groups. Khorasani 2009 which used an intra‐individual design reported zero events and therefore a sensitivity analysis to explore the impact of including it in the analysis is not required. Akhtar 1996 reported data on the grade of infection, which is reported in Table 11 but is not further analysed.

Analysis

Comparison 3 Aloe vera vs topical antibiotics, Outcome 2 Infection (time points between 14 days and 2 months).

Secondary outcome: pain

One study reported the mean reductions in pain scores from baseline (Panahi 2012) and another reported time taken to achieve pain‐free status; data from this study were reported differently between the groups and are presented in Table 12 but not analysed further (Shahzad 2013). The data from Panahi 2012 suggest that there is probably a slightly greater decrease in pain in the Aloe Vera group (mean decrease from baseline 5.68) compared with the SSD group (mean decrease from baseline 4.54). The difference in means was 1.14 (95% CI 0.02 to 2.26) (Table 12). This is moderate certainty evidence downgraded once due to imprecision.

Secondary outcome: costs

One study reported data on the cost per percentage of TBSA healed but with no measures of variance (Shahzad 2013). These data are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Summary of comparison

It is uncertain whether there is a difference in the mean number of days to healing between Aloe Vera and topical antibiotics (very low certainty evidence, downgraded due to detection bias and imprecision). It is unclear whether Aloe Vera may change the proportion of burns healed at 26 days compared with SSD (low certainty evidence, downgraded twice for imprecision). It is uncertain whether there is a difference in the incidence of infection between the groups (very low certainty evidence, downgraded due to detection bias and imprecision). Table 3.

4. Iodine‐based treatments versus topical antibiotic (2 studies, 158 participants)

Two studies compared an iodine‐based treatment with SSD (Homann 2007; Li 1994). Li 1994 was a four‐armed study that compared 0.25% iodophor with SSD, and also included groups treated with ethacridine lactate (Rivanol) and moist burn ointment (see comparisons 8, 12, 16 and 17). There were 115 participants (aged over 16 years) with injuries described as deep second‐degree burns between 1% to 12% TBSA in the trial, of whom 62 were in groups relevant to this comparison. Homann 2007 used an intra‐individual study design and compared 3% pyrrolidone iodine liposome hydrogel (Repithel) with SSD (10 mg/g) in 43 participants with a mean TBSA of 11%; their age was not reported.

Primary outcome: wound healing

Table 4

Both Homann 2007 and Li 1994 reported the mean time to healing of all wounds; this showed that the effect of iodine was very uncertain. The pooled difference in means was ‐0.47 days (95% CI ‐2.76 to 1.83; I2 = 42%) Analysis 4.1; this is very low certainty evidence, downgraded once due to risk of detection bias for the participants in Homann 2007 and twice due to imprecision; very wide confidence intervals included both the possibility of longer (by 2.8 days) or shorter (by 1.8 days) healing for participants in the iodine antiseptic group. It was not clear whether the analysis accounted for the paired data. Because of the intra‐individual design used by Homann 2007 we also report separately the effect estimate for this study (MD ‐1.40 days, 95% CI ‐3.39 to 0.59) and for Li 1994 (MD 1.00 days; 95% CI ‐1.98 to 3.98).

Analysis

Comparison 4 Iodine‐based treatments versus topical antibiotics, Outcome 1 Wound healing (mean time to healing).

Primary outcome: infection

Table 4

Neither study reported data on change in infection status.

Secondary outcome: adverse events

Table 4

Homann 2007 reported data on adverse events in all participants and distinguished local events (which could be related to the different treatments given to the participants). It is uncertain whether there is a difference in incidence of adverse events between the groups; the RR was 0.86 (95% CI 0.35 to 2.10) (Table 12). This is very low certainty evidence, downgraded once due to risk of detection bias and twice due to high levels of imprecision; wide confidence intervals included both the possibility of both a 65% decrease and a 210% increase in events in the intervention group. It is also unclear whether the intra‐individual design was accounted for in the analysis.

Secondary outcome: costs

Li 1994 reported total treatment costs for each intervention group. There may be little or no difference in cost between the iodine and SSD treatments. The mean cost for the iodine group was RMB 621 compared with RMB 674 for the SSD group; the difference in means was RMB ‐53 (‐133.29 to 27.29) Table 12. This is low certainty evidence, downgraded twice due to high levels of imprecision; confidence intervals included both the possibility of substantially lower (RMB ‐133) and somewhat higher (RMB 27) costs.

Summary of comparison

The effect of iodine‐based products on would healing is very uncertain; the confidence intervals for the estimate included the possibility of both benefit and harm (very low certainty evidence, downgraded for risks of bias and imprecision). There were no evaluable data relating to infection Table 4.

5. Sodium hypochlorite versus topical antibiotics (1 study, 20 participants)

Ning 2008 compared sodium hypochlorite with SSD in 20 adult participants with deep partial‐thickness burns less than 60% TBSA. The study used an intra‐individual design and randomised comparable burns on the same person to each treatment. It was not clear whether the analysis was adjusted to take account of the paired data.

Primary outcome: wound healing

Ning 2008 reported the mean time to healing for burns treated with each intervention. Sodium hypochlorite may slightly decrease the mean time to healing. Mean time to healing for burns treated with sodium hypochlorite was 20 days compared with 22 days for burns treated with SSD (MD ‐2.10 days; 95% CI ‐3.87 to ‐0.33) Analysis 5.1. This is low certainty evidence because of the very high levels of imprecision (Table 10; Table 13). The confidence intervals were very fragile due to the small number of participants and uncertainty as to whether the paired data were correctly analysed.

5

Summary of evidence and GRADE judgements for comparisons/outcomes with sparse data

Analysis

Comparison 5 Silver‐based antiseptics versus non‐antimicrobial, Outcome 1 Wound healing (mean time to healing).

Primary outcome: infection

Ning 2008 did not report data on change in infection status.

Secondary outcome: adverse events

Ning 2008 reported that there were no serious adverse events in either treatment group. Total adverse events were not reported. Low certainty evidence, downgraded twice for serious imprecision.

Summary of comparison

Sodium hypochlorite may slightly decrease the mean time to wound healing compared with SSD. This is low certainty evidence, downgraded twice due to imprecision. There were no analysable data for infection.

6. Chlorhexidine or polyhexanide (biguanides) versus topical antibiotics (2 studies, 115 participants)

Piatkowski 2011 randomised 72 burns from 60 adult participants to SSD or a polyhexanide‐containing dressing.

Thomas 1995 was a three‐armed study that compared chlorhexidine‐containing dressing with SSD. A third group were treated with a non‐antimicrobial dressing (see comparison 15). Fifty adults and children with a total of 54 burns were randomised; 34 of these burns were treated in groups relevant to this comparison; all burns were described as minor and the mean TBSA was less than 1% in all groups. In both studies it was unclear whether the analyses correctly adjusted for the design of the study with multiple burns from some participants.

Primary outcome: wound healing

Piatkowski 2011 and Thomas 1995 both reported the mean time to wound healing for each group but with no measure of variance. These data are reported in Table 10 but are not further analysed; no GRADE assessment was possible.

Primary outcome: infection

Neither study reported data on infection but Thomas 1995 reported the proportion of wounds with bacteria and pathogenic bacteria at baseline and post treatment; this is noted in Table 11 but the data are not extracted or analysed; no GRADE assessment was possible.

Secondary outcome: pain

Piatkowski 2011 reported pain at dressing change and between dressing changes at a number of time points from baseline up to 14 days. None of these data had any measure of variance so are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Secondary outcome: costs

Piatkowski 2011 reported costs per day for materials and personnel, and total costs, but without measures of variance. Again these data are shown in Table 12 but are not further analysed; no GRADE assessment was possible.

Summary of comparison

There were no analysable data for either of the primary outcomes or any secondary outcome.

7. Octenidine versus topical antibiotics (1 study, 30 participants)

Radu 2011 used an intra‐individual design with 30 adult participants with injuries described as second‐degree, partial‐thickness burns more than 3% TBSA to compare octenidine with SSD. It was unclear whether the analyses reported took the intra‐individual design into account.

Primary outcome: wound healing

Radu 2011 did not report wound healing.

Primary outcome: infection

Radu 2011 did not report change in infection status.

Secondary outcome: pain

Radu 2011 reported that the median VAS for the octenidine group was 3 (range 1 to 6) compared with 6 in the SSD group (range 3 to 8). Mean scores were not reported and these data were not analysed further (Table 12); no GRADE assessment was possible.

Summary of comparison

There were no analysable data for either of the primary outcomes or any secondary outcome.

8. Ethacridine lactate (Rivanol) versus topical antibiotics (1 study, 115 participants)

Li 1994 was a four‐armed study that compared ethacridine lactate (Rivanol) with SSD, and also included groups treated with iodophor and moist burn ointment (see comparisons 4, 12, 16 and 17). There were 115 participants (aged over 16 years) with injuries described as deep second‐degree burns between 1% to 12% TBSA in the trial, of whom 60 were in groups relevant to this comparison.

Primary outcome: wound healing

Li 1994 reported the mean time to healing of all wounds. There may be little or no difference between participants treated with ethacridine and those treated with SSD in mean time to healing. The difference in means was 2 days (95% CI ‐0.57 to 4.57) Analysis 6.1. This is low certainty evidence, downgraded twice due to high levels of imprecision; wide and fragile confidence intervals included both the possibility of healing being shorter by half a day or longer by over 4 days (Table 10; Table 13).

Analysis

Comparison 6 Honey versus non‐antibacterial dressing, Outcome 1 Wound healing (hazard ratio).

Primary outcome: infection

Li 1994 did not report data on change in infection status.

Secondary outcome: costs

Li 1994 reported total treatment costs for each intervention group. There may be little or no difference in costs between the ethacridine lactate and SSD groups. The mean cost per participant was RMB 598 for ethacridine lactate versus RMB 674 for SSD. The difference in means was RMB ‐76 (95% CI ‐156.34 to 4.34) (Table 12). This is low certainty evidence, downgraded twice due to serious imprecision; wide and fragile confidence intervals included both a very considerable saving (RMB 156) and a small cost (RMB 4) for the antiseptic intervention.

Summary of comparison

There may be little or no difference in time to healing between the ethacridine lactate and the SSD groups. This is low certainty evidence, downgraded twice due to imprecision. There were no data reported on infections.

9. Merbromin versus topical antibiotic (1 study, 125 participants)

Piccolo‐Daher 1990 was a five‐armed trial with 125 participants of whom 50 were relevant to this comparison between merbromin and zinc sulphadiazine. Three arms with 75 participants in total were relevant to the review (see comparison 19). Although the unit of analysis was reported to be the burn rather than the participant, it appeared that participants only presented with one burn, therefore we do not believe that there is a unit of analysis issue.

Primary outcome: wound healing

Piccolo‐Daher 1990 reported the mean time to wound healing. Merbromin may slightly decrease the mean time to healing compared with zinc sulphadiazine. Mean time to healing was 11.32 days in the merbromin group compared with 14.8 days in the zinc sulfadiazine group. The difference in means was ‐3.48 (95% CI ‐6.85 to ‐0.11). This is low certainty evidence with wide, fragile confidence intervals, downgraded twice due to high levels of imprecision (Table 10; Table 13).

Primary outcome: infection

Piccolo‐Daher 1990 did not report data on change in infection status.

Summary of comparison

Merbromin may slightly decrease the mean time to healing compared with zinc sulphadiazine (low certainty evidence); there were no data reported on infections.

10. Arnebia euchroma versus topical antibiotic (1 study, 49 participants)

Nasiri 2016 was an intra‐individual design trial that randomised burns on 49 participants to the herbal extract of A euchroma or SSD.

Primary outcome: wound healing

Nasiri 2016 reported mean time to healing and the number of healing events at multiple time points, so we were able to calculate an HR. It is unclear whether there is a difference in the 'chance' of healing over time between treatment with A euchroma or SSD; this is uncertain as fragile confidence intervals spanned both benefit and harm. The HR was 1.42 (95% CI 0.91 to 2.21). There may be a small difference (3.6 days) in the mean time to healing (95% CI ‐6.41 to ‐1.06). In both cases this is low certainty evidence, downgraded twice for imprecision. In both analyses it was unclear whether correct adjustment for the intra‐individual design was undertaken; this increases uncertainty around the estimates of effect.

Primary outcome: infection

Nasiri 2016 reported the numbers of burns with an infection score between 0 and 5 for each treatment; one point was awarded for each symptom of infection. These data are reported in Table 11 but are not further analysed; no GRADE assessment was possible.

Secondary outcome: adverse events

Nasiri 2016 reported scores for specific complications such as burning, pain, itching, warming, and incidence of allergic reactions and requiring skin grafts. It was not clear that these represented data on the number of burns with associated adverse events in each group. The data are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Secondary outcome: pain

Nasiri 2016 reported pain scores graphically at multiple time points after injury (days) and at multiple time points after dressing (minutes). We could not extract confidence intervals from the graphs but all were reported by study authors to have differences between groups (P reported < 0.05). The data are noted in Table 12 but are not further analysed; no GRADE assessment was possible.

Summary of comparison:

It is unclear whether there is a difference in time to healing between treatment with A euchroma or SSD assessed by the hazard or 'chance' of healing over time. There may be a small reduction in the mean time to healing in burns treated with Aeuchroma compared with those treated with SSD. In both cases this is low certainty evidence. There were no evaluable data on the incidence of infection.

Comparisons between two antiseptics

11. Chlorhexidine versus povidone iodine (1 study, 213 participants)

Han 1989 enrolled 213 participants with burns less than 10% TBSA; approximately 20% were children. Participants were randomised to Bactigras (tulle‐gras wide‐meshed gauze dressing impregnated with 0.5 per cent chlorhexidine acetate BP) or Inadine (synthetic rayon dressing impregnated with 10 per cent povidone iodine ointment).

Primary outcome: wound healing

Han 1989 reported mean time to wound healing. There may be a slightly increased mean time to healing in the chlorhexidine group. Mean time to healing was 11.69 days in the chlorhexidine group compared with 9.48 in the iodine group. The difference in means was 2.21 days (95% CI 0.34 to 4.08) Analysis 8.1. This is low certainty evidence, downgraded once due to risk of reporting bias and once due to imprecision because of wide confidence intervals (Table 10; Table 13).

Analysis

Comparison 8 Iodine‐based antiseptics versus non‐antibacterial treatments, Outcome 1 Wound healing (mean time to healing).

Primary outcome: infection

Han 1989 reported incident infections in each group. It is uncertain whether there is a difference in infection incidence between the treatments. There were 4/102 in the chlorhexidine and 4/111 in the povidone iodine groups. The RR was 1.09 (95% CI 0.28 to 4.24). This is very low certainty evidence, downgraded once due to risk of reporting bias and twice due to very serious imprecision (Table 11; Table 13).

Secondary outcome: pain

Han 1989 reported mean pain at rest and at dressing removal using a VAS. It is uncertain whether there is a difference between the chlorhexidine and povidone iodine groups on either measure of pain. The mean VAS score for pain at rest was 11.44 in the chlorhexidine group and 9.18 in the povidone iodine group. The difference in means was 2.26 (95% CI ‐2.26 to 6.78). The mean score for pain at dressing change was 8.75 in the chlorhexidine group and 6.66 in the povidone iodine group. The difference in means was 2.09 (95% CI ‐2.00 to 6.18) (Table 12). In both cases this is very low certainty evidence, downgraded once for risk of reporting bias and twice for very serious imprecision; wide confidence intervals included the possibility of both lower (‐2) and much greater pain scores (+6) in the chlorhexidine group.

Secondary outcome: resource use

Han 1989 reported the mean number of hospital visits for each participant in the two treatment groups. There may be little or no difference between the chlorhexidine (mean 2.64) and iodine (mean 3.03) groups. The difference in means was 0.25 visits (95% CI ‐ 0.02 to 0.52) (Table 12). This is low certainty evidence, downgraded once due to risk of reporting bias and once due to imprecision, as the confidence intervals included both the possibility of both slightly fewer and somewhat more visits in the intervention group.

Summary of comparison

Chlorhexidine‐based dressings may result in a slightly longer mean time to healing than povidone iodine dressings (low certainty evidence, downgraded due to risk of bias and imprecision). It is uncertain whether there is a difference between chlorhexidine and povidone iodine in the number of incident infections in burn wounds (very low certainty evidence, downgraded due to risk of bias and serious imprecision).

12. Iodophor versus ethacridine lactate (1 study, 115 participants)

Li 1994 was a four‐armed study that compared 0.25% iodophor with ethacridine lactate, and also included groups treated with SSD and moist burn ointment (see comparisons 4, 8, 16 and 17). There were 115 participants (aged over 16 years) with injuries described as deep second‐degree burns between 1% to 12% TBSA in the trial, of whom 53 were in groups relevant to this comparison.

Primary outcome: wound healing

Li 1994 reported mean time to wound healing. There may be little or no difference in healing time between participants treated with iodophor and those treated with ethacridine lactate. Mean time to healing was 31 days in the iodophor group compared with 32 days in the ethacridine lactate group (MD ‐1.00 day (95% CI ‐4.31 to 2.31) (Table 10; Table 13). This is low certainty evidence due to high levels of imprecision; wide confidence intervals included the possibility of both a shorter healing time by 4 days and a longer healing time by 2 days for the iodine group.

Primary outcome: infection

Li 1994 did not report data on change in infection status.

Secondary outcome: costs

Li 1994 reported total treatment costs for each intervention group. There may be little or no difference in costs between the iodine and ethacridine‐lactate groups. The cost per participant was RMB 621 in the iodine group compared with RMB 598 in the ethacridine‐lactate group. The difference in means was RMB 23.00 (95% CI ‐51.07 to 97.07) (Table 12). This is low certainty evidence, downgraded twice due to high levels of imprecision resulting from small numbers of participants; wide confidence intervals included both the possibilities of a saving of RMB 51and an additional cost of RMB 97.

Summary of comparison

There may be little or no difference in mean time to healing for participants treated with iodophor or ethacridine lactate (low certainty evidence). There were no data on infection.

Comparisons between antiseptics and treatments without antimicrobial properties

13. Silver dressings versus non‐antimicrobial treatments or no treatment (3 studies, 299 participants)

Chen 2006 was a three‐armed trial that randomised 191 participants with burns described as being second‐degree to a silver nanoparticle dressing or to Vaseline gauze. A third group of participants were treated with SSD (see comparison 1). The number of participants in groups relevant to this comparison was 128. Jiao 2015 randomised 76 participants with partial‐thickness burns to nanocrystalline silver or Vaseline gauze; in each case the dressing was applied over human epidermal growth factor. Healy 1989 randomised 32 participants (mostly children) to silver‐impregnated porcine xenograft or paraffin gauze.

Primary outcome: wound healing

Table 5

Healy 1989 reported the proportion of wounds completely healed in each group by 14 days. There may be little or no difference between silver xenograft and paraffin gauze in proportion of wounds healed; 9/16 wounds healed in the silver group compared with 8/16 in the control group. The RR was 1.13 (95% CI 0.59 to 2.16) (low certainty evidence, downgraded twice because of serious imprecision) Table 10. The mean time to healing of these wounds was also reported but as not all wounds healed these data are reported in Table 10 but are not further analysed.

Chen 2006 and Jiao 2015 reported the mean time to healing for all wounds. The mean time to healing is probably slightly shorter in the silver group compared with the gauze group: ‐3.49 days (95% CI ‐4.46 to ‐2.52; I2 = 0%) a reduction from 15.87 days in the gauze group to 12.38 days. This is moderate certainty evidence downgraded because of imprecision.

Primary outcome: infection

Table 5

It is very uncertain whether there is a difference in wound infections between silver and non‐antimicrobial treatments. Jiao 2015 reported the proportion of wounds testing positive for bacteria at 21 days as 1/38 in the silver group and 8/38 in the gauze group. The RR was 0.13 (95% CI 0.02 to 0.95). However this is not a measure of clinical infection. Healy 1989 reported data on specific bacteria colonisation but otherwise reported only that there was no difference in the infection rate between the groups; these data are noted in Table 11. This is very low certainty evidence, downgraded once for indirectness and twice for imprecision.

Secondary outcomes

Neither Chen 2006 nor Healy 1989 reported data on any secondary outcome while Jiao 2015 reported only one specific type of adverse event (scar hyperplasia); this is noted in Table 12 but is not further analysed.

Summary of comparison

Silver xenograft may make little or no difference to the proportion of burn wounds that heal by 14 days compared to a non‐antimicrobial (paraffin gauze) dressing (low certainty evidence downgraded twice for imprecision). Silver nanoparticle dressings probably result in burns healing in a slightly shorter mean time compared with Vaseline/petroleum gauze (moderate certainty evidence downgraded for imprecision). It is very uncertain whether there is any difference between the dressings in infection rates (very low certainty evidence downgraded for indirectness and imprecision) Table 5.

14. Honey or honey‐based dressings versus non‐antimicrobial treatments (3 studies, 256 participants)

Subrahmanyam 1993b randomised 92 participants to honey‐impregnated gauze or a bio‐occlusive, moisture‐permeable polyurethane dressing. Subrahmanyam 1994 randomised 64 participants to honey‐impregnated gauze or amniotic membrane; no other details of the comparison dressing were given. Subrahmanyam 1996a randomised 100 participants to honey plus dry guaze or an autoclaved potato peeling dressing plus dry gauze. All studies enrolled a mixture of adults and children, although most participants were adults where reported. All three studies enrolled participants with partial‐thickness burns less than 40% TBSA. Subrahmanyam 1996b compared honey with a mixed standard treatment group in which the following treatments were used: soframycin (90 participants), Vaseline‐impregnated gauze (90 participants), OpSite (90 participants), sterile gauze (90 participants) or left exposed (90 participants).

Primary outcome: wound healing

Table 6

Two trials reported data at numerous time points allowing hazard ratios for healing to be calculated (Subrahmanyam 1994, Subrahmanyam 1996a). Honey probably somewhat increases the 'chance' of healing over time for partial‐thickness burns compared with non‐anti‐microbial treatments. The pooled HR was 2.86 (95% CI 1.60 to 5.11; I2 = 50%) Analysis 6.1. This is moderate certainty evidence downgraded once for imprecision. All wounds healed in both groups over the total time period assessed, but the hazard ratio corresponds to the increased likelihood of healing at earlier time points in the honey groups. As HRs could be calculated for all trials with dichotomous data, we have not reported RRs.

All four trials also reported mean times to healing, or data were available as a result of contact with the study author by a previous Cochrane Review (Jull 2015). On average, honey results in a somewhat shorter mean time to healing compared with the non‐antibacterial dressings evaluated. The pooled difference in means was ‐5.32 days (95% CI ‐6.30 to ‐4.34; I2 = 71%) Analysis 6.2. This is high certainty evidence although some of the comparators used are atypical.

Analysis

Comparison 6 Honey versus non‐antibacterial dressing, Outcome 2 Wound healing (mean time to healing).

Primary outcome: infection

Table 6

Three trials reported a measure of infection. However this was based on swab cultures which are only an indirect measure of infection and do not correspond to clinical infections. Subrahmanyam 1993b reported incidence of infection on day 8 in both groups. It is uncertain whether there may be fewer incidences of infection in wounds treated with honey compared with polyurethane dressing. There were 8/46 infections reported in the honey group compared with 17/46 in the polyurethane dressing group. The RR was 0.47 (95% CI 0.23 to 0.98) (very low certainty evidence, downgraded twice for imprecision and twice for indirectness).

Subrahmanyam 1994 and Subrahmanyam 1996a reported incidence of persistent positive swab cultures at day 7. It is uncertain whether persistent infections differ in participants treated with honey compared with participants treated with comparator topical treatments. The total number of participants considered to have a persistent infection was 8/78 in the honey groups compared with 53/69 in the non‐antibacterial groups.The pooled RR for persistent infection at day 7 was 0.15 (95% CI 0.06 to 0.40; I2 = 50%) Analysis 6.3. This is very low certainty evidence, downgraded twice for indirectness and once for imprecision.

Analysis

Comparison 6 Honey versus non‐antibacterial dressing, Outcome 3 Persistent positive swabs (up to 30 days).

Secondary outcome: adverse events

Table 6

Subrahmanyam 1993b and Subrahmanyam 1994 reported information on some adverse events but it was not clear that these represented all reported adverse events. Subrahmanyam 1996a reported that there were no adverse events in either the honey or the SSD group. This is very low certainty evidence, downgraded twice for serious imprecision and indirectness.

Secondary outcome: pain

Subrahmanyam 1994 measured pain using a four‐point scale and reported the number of participants in each group with no or mild pain and with moderate or severe pain. The mean scores were not reported and these data are not further analysed. Subrahmanyam 1996a reported only that subjective relief of pain was the same in both treatment groups. These data are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Summary of comparison

Based on the hazard or 'chance' of healing over time, honey probably, on average, somewhat shortens the time to healing for partial‐thickness burns compared with a range of non‐antibacterial alternatives, including treatments not commonly used in clinical practice. There is high certainty evidence of some reduction in the mean time to wound healing in the wounds treated with honey. It is uncertain if burns treated with honey may develop fewer infections than those treated with the comparison treatments. Table 6.

15. Chlorhexidine (biguanide) versus non‐antimicrobial treatments (5 studies, 516 participants)

Five studies compared chlorhexidine with no treatment or a non‐antimicrobial treatment. Inman 1984 randomised 121 participants to SSD plus chlorhexidine versus SSD alone in participants with full‐thickness burns; full‐thickness injuries were less than 15% TBSA. Other studies used chlorhexidine‐impregnated paraffin gauze or tulle‐gras. Neal 1981 (51 participants), Phipps 1988 (196 participants) and Thomas 1995 (50 participants) enrolled people with burns less than 5% TBSA while Wright 1993 (98 participants) required that burns be suitable for outpatient treatment. Comparators were plastic film dressing (Neal 1981) or hydrocolloid dressing (Phipps 1988; Thomas 1995; Wright 1993). Where reported, all studies enrolled a mix of adults and children. Thomas 1995 also assigned participants to a third arm treated with SSD (see comparison 6). As previously noted, some participants in Thomas 1995 had multiple burns analysed in the study, creating unit of analysis issues.

Primary outcome: wound healing

Table 7

Neal 1981 reported the number of participants whose wounds healed at multiple time points and presented a Kaplan‐Meier curve. The trial did not show a clear difference between chlorhexidine and non‐antimicrobial film dressing in time to wound healing; wide and fragile CIs spanned both benefit and harm so a clear treatment effect is not apparent. The calculated HR, based on 25 participants in the chlorhexidine group and 26 in the film‐dressing group, was 0.71 (95% CI 0.39 to 1.29) (Table 10). Neal 1981 also reported the mean time to healing. This indicated that the mean time to healing may be slightly longer (14 days) in the chlorhexidine group (compared with 10 days in the film dressing group) with a difference in means of 4.08 days (95% CI 0.73 to 7.43); again the estimate was imprecise. Both Phipps 1988 and Thomas 1995 reported the mean time to healing in each group but did not report a measure of variance. Wright 1993 reported the median time to wound healing in each group. The data for these three trials are reported in Table 10 but are not further analysed. All the evidence is low certainty, downgraded twice because of serious imprecision due to low participant numbers, wide confidence intervals and poor reporting. The three trials that did not present analysable data were also all at high risk of bias across one or more domains. Because the study that had unit of analysis issues did not contribute to the analysis, (Thomas 1995) no sensitivity analysis was undertaken.

Primary outcome: infection

Table 7

Inman 1984 reported the numbers of participants in each group with infection. It appeared that there were a number of post‐randomisation exclusions from the study, numbers are reported on a completed case basis. Neal 1981 reported the proportion of participants in each group with proven infections. It is uncertain whether there is a difference between the treatments. On average the RR for wound infection from these two studies was RR 1.11 (95% CI 0.54 to 2.27; I2 = 0%) Analysis 7.2. This is very low certainty evidence, downgraded twice due to high levels of imprecision and once due to attrition bias in Inman 1984: wide confidence intervals included the possibility of both substantial benefits and harms associated with the intervention.

Analysis

Comparison 7 Chlorhexadine versus non‐antibacterial dressing, Outcome 2 Infection (up to 30 days).

Phipps 1988 reported proportions of participants with specific wound flora and Thomas 1995 reported percentages of wounds with bacteria and pathogenic bacteria; these data are noted in Table 11 but are not further analysed.

Secondary outcome: adverse events

Wright 1993 reported the number of participants with an adverse event in each group. It is uncertain whether chlorhexidine decreases the number of people experiencing adverse events. In the chlorhexidine group, 1/49 participants experienced an adverse event, compared with 5/49 in the comparison group. The RR was 0.20 (95% CI 0.02 to 1.65) (Table 12). This is very low certainty evidence, downgraded once due to risk of detection bias, once due to attrition bias and twice due to very high levels of imprecision as a result of very wide confidence intervals, which included a possible 98% reduction and also a 65% increase in events associated with the antiseptic intervention.

Secondary outcome: pain

Inman 1984 reported the numbers of participants in each group with pain sufficient to stop treatment. Neal 1981 reported qualitatively that chlorhexidine treatment was perceived as painful. Wright 1993 reported summing the VAS for each visit; the scores were not reported but a P value was given. All these data are reported in Table 12 but are not further analysed; no GRADE assessment was possible.

Secondary outcome: mortality

Inman 1984 reported total and infection‐related mortality in each treatment group. It is uncertain whether chlorhexidine in addition to SSD alters mortality. A total of 3/54 people died in the chlorhexidine group compared with 4/67 in the SSD‐alone group. The RR was 0.93 (95% CI 0.22 to 3.98) (Table 12). This is very low certainty evidence, downgraded once for attrition bias and twice for very serious imprecision due to wide confidence intervals, which included a possible 78% reduction and an almost 400% increase in deaths.

Secondary outcome: resource use

Wright 1993 reported the number of dressings used in each group as 2.8 in the chlorhexidine group and 2.61 in the hydrocolloid group. No measures of variance were reported and the data were not further analysed but are shown in Table 12; no GRADE assessment was possible.

Summary of comparison

Despite being evaluated in multiple trials the evaluable data were limited. There may be little or no difference in the time to wound healing between chlorhexidine and a film dressing (low certainty evidence downgraded twice for imprecision). It is uncertain whether the use of chlorhexidine reduces the incidence of infection compared with no additional antibacterial treatment (very low certainty evidence, downgraded twice due to imprecision and once due to attrition bias). It is also uncertain whether use of chlorhexidine plus SSD reduces mortality compared with SSD alone (very low certainty evidence, downgraded twice for imprecision and once for risk of attrition bias). Table 7

16. Iodine‐based treatments versus non‐antimicrobial treatments/no intervention (4 studies, 663 participants)

Carayanni 2011 randomised 217 participants with superficial or deep partial‐thickness thermal burns less than 15% TBSA to povidone iodine with a barrier of bepanthenol cream or MEBO. Randomisation was stratified by burn depth. Li 1994 was a four‐armed study that compared iodophor with moist burn ointment and also included groups treated with SSD and ethacridine lactate (see comparisons 4, 8, 12 and 17). There were 115 participants (aged over 16 years) with injuries described as deep second‐degree burns between 1% to 12% TBSA in the trial, of whom 55 were in groups relevant to this comparison. Li 2006 randomised 277 participants with superficial, deep or residual burn wounds to iodine gauze or to carbon fibre dressing. Yang 2013 enrolled 60 participants with residual burn wounds after one month of treatment and randomised burn areas to iodophor gauze or to a hydrogel dressing; this was an intra‐individual design.

Primary outcome: wound healing

Table 8

Yang 2013 reported the proportion of wounds healed at seven and 14 days. Iodophor gauze may reduce the chances of residual burn wounds healing after 14 days: RR was 0.17 (95% CI 0.08 to 0.34) (Table 10). It was unclear whether the analysis adjusted for the paired data from the intra‐individual design.This is low certainty evidence, downgraded twice for imprecision due to uncertainties about the analysis and small numbers of participants.

Li 1994 reported mean time to wound healing as did Li 2006. It is unclear whether the use of iodine reduces the mean time to healing because the certainty of the evidence is very low. Clinical differences in the treatments used and very high levels of statistical heterogeneity (I2 = 99%) meant that pooling was unlikely to produce a meaningful answer. Li 1994 reported that mean time to healing for wounds was 31 days in the iodophor group and 57 days in the MEBO group (Li 1994), with a difference in means of ‐26 days (95% CI ‐30.48 to ‐21.52). Li 2006 reported that mean time to healing for wounds was 20.67 days in the iodine‐gauze group compared with 15.29 days in the carbon‐dressing group, with a difference in means of 5.3 days (95% CI 3.09 to 7.67) Analysis 8.1. This is very low certainty evidence, downgraded twice for inconsistency and twice for imprecision.

Primary outcome: infection

Table 8

Carayanni 2011 reported the numbers of participants with infection. There may be little or no difference between iodine and MEBO in the incidence of infections. There were 8/107 participants with infections in the iodine group compared with 6/104 in the MEBO group. The RR was 1.30 (95% CI 0.47 to 3.61). This is low certainty evidence, downgraded twice for very high levels of imprecision with wide confidence intervals, which included the possibility of a both a reduction of 53% and an increase of 360% in infection rates for the iodine intervention (Table 11).

Yang 2013 reported bacterial presence in wounds and stated that there was no evidence of a difference between the groups (Table 11); these data were not further analysed.

Secondary outcome: adverse events

Table 8

Carayanni 2011 reported adverse events including infections. There may be little or no difference between iodine and MEBO in the incidence of adverse events. There were 8/107 participants with reported events (all infections) in the iodine group and 11/104 in the MEBO group. The RR was 0.71 (95% CI 0.30 to 1.69). This is also low certainty evidence, downgraded twice due to the very high levels of imprecision with wide confidence intervals, which included the possibility of a 70% reduction or a 70% increase in events for the iodine intervention (Table 12).

Secondary outcome: pain

Yang 2013 reported pain at dressing change as the number of wounds and the level of pain. Carayanni 2011 reported median pain scores in graphical form only. In both cases these data are noted in Table 12 but are not extracted or analysed further; no GRADE judgement was possible.

Secondary outcome: resource use

Carayanni 2011 reported reduction in length of hospital stay from an expected duration based on burn characteristics. Hospital stay is probably reduced by slightly less time in participants treated with iodine compared with those treated with MEBO. There was a reduction of 3.01 days in the iodine group compared with 3.63 in the MEBO group; the difference in means was 0.62 days (95% CI 0.05 to 1.19) (Table 12). This is moderate certainty evidence, downgraded once for imprecision due to low numbers of participants.

Secondary outcome: costs

Li 1994 and Carayanni 2011 reported total treatment costs for each intervention group. It is uncertain whether iodine‐based treatments reduce costs compared with MEBO. Clinical differences in the treatments used and very high levels of statistical heterogeneity (I2 = 99%) meant that pooling was unlikely to produce a meaningful answer (Analysis 8.2). Li 1994 reported costs of RMB 621 for the iodophor group compared with RMB 1836 for the MEBO group (difference in means RMB ‐1215, 95% CI ‐1412.96 to ‐1017.04). Carayanni 2011 reported costs of EUR 566 for povidone iodine and EUR 529 for the MEBO group (difference in means EUR 36.55, 95% CI ‐7.33 to 80.43). This is very low certainty evidence, downgraded twice for high levels of inconsistency and twice for imprecision due to low participant numbers and wide confidence intervals.

Analysis

Comparison 8 Iodine‐based antiseptics versus non‐antibacterial treatments, Outcome 2 Costs (duration 18 days +).

Summary of comparison

It is uncertain whether iodine‐based treatments decrease or increase the mean time to healing compared with treatments without antibacterial properties (very low certainty evidence downgraded twice due to inconsistency and twice for imprecision). Iodophor gauze may reduce the chances of residual burn wounds healing within 14 days compared with hydrogel treatment (low certainty evidence downgraded twice due to imprecision). There may be little or no difference in the occurrence of either infections or adverse events including infections, between povidone iodine and a non‐antibacterial comparator (low certainty evidence downgraded twice due to imprecision) Table 8.

17. Ethacridine lactate versus non‐antimicrobial treatments (1 study, 115 participants)

Li 1994 was a four‐armed study that compared ethacridine lactate with moist burn ointment and also included groups treated with SSD and iodophor (see comparisons 4, 8, 12 and 16). There were 115 participants (aged over 16 years) with injuries described as deep second‐degree burns between 1% to 12% TBSA in the trial, of whom 46 were in groups relevant to this comparison.

Primary outcome: wound healing

Li 1994 reported mean time to wound healing. Mean time to wound healing may be reduced in the ethacridine lactate group compared with the MEBO group. Mean times to heal were 32 days for the ethacridine group and 57 days for the MEBO group; the difference in means was ‐25 days (95% CI ‐29.21 to ‐20.79). This is low certainty evidence downgraded twice due to high levels of imprecision resulting from small numbers of participants (Table 10; Table 13).

Primary outcome: infection

Li 1994 did not report data on change in infection status.

Secondary outcome: costs

Li 1994 reported total mean treatment costs for each intervention group. Total costs may be lower in the ethacridine lactate group compared with the MEBO group. Costs in the ethacridine group were RMB 598 compared with RMB 1836 for people in the MEBO group. The difference in means was RMB ‐1238 (95% CI ‐1435.98 to ‐1040.22) Table 12. This is low certainty evidence downgraded twice due to high levels of imprecision resulting from small numbers of participants.

Summary of comparison

There may be a shorter mean time to healing in burns treated with ethacridine lactate compared with MEBO. This is low certainty evidence due to high levels of imprecision and fragile confidence intervals. There were no data on infection.

18. Cerium nitrate + topical antibiotic versus topical antibiotic alone (2 studies, 214 participants)

Oen 2012 randomised 154 adults with facial burns to cerium nitrate plus SSD or SSD alone while De Gracia 2001 randomised 60 participants with full or partial‐thickness burns to the same interventions.

Primary outcome: wound healing

Oen 2012 reported the median time to healing and interquartile range in each group for those participants who did not need to have surgery; data were therefore not included for all wounds and the data are not analysed further. De Gracia 2001 reported the mean time to healing for partial‐thickness burn areas (these made up part of the wound for all except one participant) but not data for whole wounds, as the full‐thickness burn areas were grafted when ready. These data are therefore not analysed further and no GRADE assessment was possible (Table 10).

Primary outcome: infection

De Gracia 2001 found that the effect of treatment with cerium nitrate in addition to SSD is unclear in terms of the number of participants with sepsis at up to five days and then subsequently compared with SSD alone. In total 1/30 participants in the cerium nitrate group had sepsis versus 4/30 in the control group. The RR was 0.25 (95% CI 0.03 to 2.11) so the wide confidence intervals included the possibility of both benefit and harm. This was also the case for the number of participants with post‐treatment infection compared with SSD alone; 3/30 participants developed an infection in the cerium nitrate group compared with 6/30 in the control group (RR 0.50, 95% CI 0.14 to 1.82). In both cases this is low certainty evidence, which was downgraded twice due to serious imprecision because of low numbers of events and participants (Table 11; Table 13).

Oen 2012 did not report data on change in infection status.

Secondary outcome: pain

Oen 2012 reported mean pain scores both generally and for procedures. Cerium nitrate plus SSD probably slightly reduces overall pain scores. In the cerium nitrate group the mean score was 0.6 compared with 1.2 in the control group. The difference in means was ‐0.60 (95% CI ‐0.70 to ‐0.50) Table 12. This is moderate certainty evidence downgraded once for imprecision due to the small number of participants.

Secondary outcome: mortality

Both De Gracia 2001 and Oen 2012 reported the number of participants who died in each treatment group. Cerium nitrate plus SSD may reduce mortality compared with SSD alone.There were 2/108 deaths in the cerium nitrate group compared with 9/106 in the SSD‐alone group. The RR was 0.22 (95% CI 0.05 to 0.99; I2 = 0%) Analysis 9.1. This is low certainty evidence downgraded twice due to imprecision because of wide confidence intervals, which were fragile due to low numbers of both events and participants. In one of the trials (Oen 2012) deaths occurred during the enrolment process so the effect of treatment group is unclear.

Analysis

Comparison 9 Cerium nitrate versus non antibacterial treatment, Outcome 1 Mortality (short‐term or unclear).

Secondary outcome: resource use

De Gracia 2001 reported the mean length of hospital stay. It is unclear whether cerium nitrate in addition to SSD reduces hospital stay. The mean length of stay was 23.3 days in the cerium nitrate group versus 30.7 days in the control group. The difference in means was ‐7.4 days (95% CI ‐16.49 to 1.69) Table 12. This is low certainty evidence downgraded twice due to imprecision because of wide confidence intervals, which included both a substantial benefit (16.5 days) for the cerium nitrate group and a small benefit (1.7 days) for the comparison group.

Summary of comparison

There were no analysable data on wound healing. The effect of cerium nitrate in addition to SSD on rates of infection and of sepsis is unclear, compared with SSD alone (low certainty evidence with wide confidence intervals including both benefit and harm, downgraded twice for imprecision). There may be lower mortality rates in the cerium nitrate group compared with the group treated with SSD alone (low certainty evidence, downgraded twice due to serious imprecision).

19. Merbromin versus sodium salicylate (1 study, 125 participants)

This comparison was addressed by one trial. Piccolo‐Daher 1990 was a five‐armed trial with 125 participants of whom 50 were relevant to this comparison. Three arms with 75 participants in total were relevant to the review (see comparison 9). As above, although the unit of analysis was reported to be the burn rather than the participant it appeared that participants only presented with one burn, therefore we do not believe that there is a unit of analysis issue.

Primary outcome: wound healing

Piccolo‐Daher 1990 found that the mean time to wound healing may be slightly reduced in participants treated with merbromin (11.32 days) compared with those treated with sodium salicylate (15.0 days). The difference in means was ‐3.68 days (95% CI ‐7.18 to ‐0.18). This is low certainty evidence downgraded twice due to imprecision (Table 10; Table 13).

Primary outcome: infection

Piccolo‐Daher 1990 did not report data on change in infection status.

Summary of comparison

Burns treated with merbromin may have a slightly shorter mean time to healing than those treated with sodium salicylate (low certainty evidence downgraded twice due to serious imprecision). There were no data on infection.

Discussion

Summary of main results

We identified 56 eligible studies with 5807 randomised participants. The majority of these assessed treatments with antiseptic properties and compared them to the topical antibiotic silver sulfadiazine. Most participants appeared to be adults, although the majority of studies enrolled both adults and children. In most studies burns were required to correspond to a classification (by the studies' authors) of second degree and to be under 40% TBSA. Some studies focused on smaller and more superficial burns and a smaller number allowed some deeper burn areas. A minority of participants had residual burn wounds, but the great majority were enrolled in the period immediately after the injury.

Wound healing

Antiseptics compared with topical antibiotics

Evidence on wound healing is mixed and largely of low certainty due to small sample sizes and wide confidence intervals.

Measuring the hazard or 'chance' of healing over time using a HR suggested that there is no clear difference in time to healing between wounds treated with silver‐based antiseptics (mainly dressings) and those treated with topical antibiotics (all SSD); this is low certainty evidence as data came from 259 participants, and the 95% CIs spanned effects of both benefit and harm for the intervention. Low certainty evidence from a larger number of studies (979 participants), reporting mean time to healing of all wounds, suggested that there may be a modest benefit of healing time, approximately three days shorter in the silver‐based antiseptics arm; while studies that reported dichotomous healing data also suggested that there may be little difference in how many wounds treated with silver antiseptics may heal by three or four weeks compared with those treated with SSD.

Measuring the hazard or 'chance' of healing over time using an HR suggested that wounds treated with honey probably have a somewhat shorter time to healing than wounds treated with topical antibiotics (moderate certainty evidence based on 580 participants). There may, on average, be a greater number of healing events measured at short term (21‐day follow‐up) (low certainty evidence). It is uncertain whether there is a difference in mean time to healing (very low certainty evidence)

It is unclear whether there is a difference in the number of healing events over a 26‐day period in burns treated with Aloe Vera compared with SSD. It is uncertain if the overall average effect of mean time to healing differs between these treatments (very low certainty evidence). There is low certainty evidence that sodium hypochlorite may be associated with a mean time to healing that was lower by around two days than for SSD. Also with low certainty evidence, there may be a small benefit (around 3.5 days) in mean time to healing from merbromin compared with zinc sulfadiazine. There is low certainty evidence that there may be a similar small benefit of around 3.6 days for treatment with extract of the herb A euchroma, which has antiseptic properties, compared with SSD, but it was unclear whether there was a difference in the 'chance' of healing over time. There is low certainty evidence that there may be little or no treatment difference in wound healing for the comparisons of ethacridine lactate or iodine‐based treatments with silver sulfadiazine. There were no usable data for the primary outcomes from trials comparing chlorhexidine, polyhexanide or octenidine to silver sulfadiazine.

Antiseptics compared with alternative antiseptics

There were few comparisons between different antiseptics. Low evidence from a single trial indicated that there may be a small benefit of around two days in mean time to healing for wounds treated with povidone iodine compared with chlorhexidine. There may be little or no difference between iodophor and ethacridine lactate in wound healing times (low certainty evidence).

Antiseptics compared with non‐antibacterial alternative treatments

Several different antiseptic agents were compared with a range of dressings without antibacterial properties. The evidence from these comparisons is generally of low certainty.

There is moderate certainty evidence, based on 204 participants in two trials that, on average, burns treated with nanocrystalline silver dressings have a slightly shorter mean time to healing (by around 3.5 days) than those treated with Vaseline gauze. There is low certainty evidence that there may be little or no difference in the number of healing events at 14 days between burns treated with silver xenograft or paraffin gauze.

Measuring the hazard or 'chance' of healing over time using a HR suggested that wounds treated with honey probably, on average, have a somewhat shorter time to healing than wounds treated with unconventional non‐antibacterial treatments, based on 164 participants treated with honey compared with amniotic membrane or potato peelings (moderate certainty evidence). There is high certainty evidence for a shorter average mean time to healing in burns treated with honey compared with non‐antibacterial treatments, including the unconventional ones assessed using the HR. Burns healed, on average, in a mean time which was 5.3 days shorter in groups treated with honey.

Comparisons involving iodine‐based treatments produced contradictory results favouring both iodine and the comparator in terms of mean time to healing; it is uncertain where the true treatment effect may lie (very low certainty evidence). There may be fewer healing events over a short follow‐up period in wounds treated with iodophor gauze compared with hydrogel (low certainty evidence). Honey was compared with a range of treatments, some of which were unconventional. There is also low certainty evidence that both merbromin and ethacridine lactate may result in shorter mean times to healing compared with a non‐antibacterial treatment.

Infection

Antiseptics compared with topical antibiotics

Most comparisons did not report data on this key outcome but the comparisons of silver and Aloe Vera with SSD showed that there may be little or no difference between the treatment arms (low certainty evidence). There is uncertainty as to the effect of treatment with honey compared with SSD on infections (very low certainty evidence).

Antiseptics compared with alternative antiseptics

It is uncertain whether there was a difference in infections between chlorhexidine and povidone iodine (very low certainty evidence). There were no data on any other comparisons between antiseptics.

Antiseptics compared with non‐antibacterial alternative treatments

It is uncertain whether there were differences in burns treated with either silver‐based dressings or honey compared with a range of non‐antimicrobial treatments, some of which were unconventional; this is very low certainty evidence in both cases. There is moderate certainty evidence of no difference in infection rates for a comparison of an iodine‐based treatment with MEBO. The only comparison that showed any evidence of a benefit in infection reduction was the use of cerium nitrate in addition to silver sulfadiazine, compared with silver sulfadiazine alone, where there is some low certainty evidence of a reduced incidence of infections and sepsis. Other comparisons did not report usable data on infection rates.

Secondary outcomes

Adverse events were not reported for many comparisons, or they were reported in qualitative terms, which made it difficult to determine event rates for each intervention group, or they were reported only for specific types of event. There may be little or no difference in adverse events for any of the comparisons that did report the number of participants with an adverse event in each group; or the impact of treatments on adverse effects is very uncertain (low or very low certainty evidence).

Mortality was low where reported and there may be little or no difference between treatment groups in most comparisons; this was based on very small numbers of events and is low certainty evidence. The exception was the comparison of cerium nitrate plus SSD with SSD alone, where there may be fewer deaths in the cerium nitrate groups; again, event rates were low and this is low certainty evidence based on 214 participants in two trials.

Pain is of particular concern to people with burns and medical personnel: often this was not reported in sufficient detail for us to analyse but there was evidence that there may be lower levels of pain in participants treated with silver dressings compared with silver sulfadiazine (low certainty evidence). Pain probably decreases slightly more from baseline in those treated with Aloe Vera compared with SSD (moderate certainty evidence). There is also some low certainty evidence that participants treated with cerium nitrate plus SSD may have less pain than those treated with SSD alone. It was uncertain whether there was a difference in pain between participants treated with chlorhexidine and those treated with povidone iodine (very low certainty evidence).

Resource use was reported for a limited number of comparisons. Frequency of dressing changes and associated implications for nursing time and costs were the most commonly reported outcomes under this heading. There was some evidence that participants treated with silver dressings may require fewer dressing changes than those treated with SSD (low certainty evidence) and some evidence that participants who are treated with honey probably have a shorter hospital stay compared with those treated with SSD (moderate certainty evidence). Reduction from an expected length of stay in hospital is probably smaller in participants treated with iodine‐based dressings compared with MEBO (moderate certainty evidence). With a few exceptions, costs were not adequately reported or showed that there may be no differences between treatment groups. In some cases there may be cost differences between groups, but this is all low certainty evidence.

Overall completeness and applicability of evidence

Although we identified a large number of studies, many of these did not report, or did not fully report, the primary outcomes of this review: wound healing and infection. Usable data on key outcomes were therefore limited and often unavailable. Only a minority of studies reported enough data to enable us to calculate the most appropriate measure of time‐to‐event data ‐ a hazard ratio. Where this was not available we were in some cases able to report a mean time to healing or a relative risk of healing for a particular time point. Neither of these measures is ideal and both may give an impression of either an effect or a lack of effect which is not truly present, particularly where the event rate is high.

Usable evidence on infection was also limited, which is disappointing in an area in which infection control is so key. Although a number of studies reported microbiological data, the proportion reporting analysable data on clinical infection was much smaller. Much of the evidence is of low certainty or very low certainty because of indirectness and imprecision.

The geographical distribution of the studies reflected the concentration of disease burden outside of Western high‐income countries. Most studies included participants described as having second‐degree burns and there were no studies focusing on full‐thickness burns, although some studies allowed participants with some area of full‐thickness (described as third‐degree) injury. Therefore any conclusions that can be drawn from this review are likely to be directly relevant only to participants with second‐degree burns limited to TBSA of 40% or lower. Their reliability for other types of burns will be reduced by indirectness.

Quality of the evidence

For most of the comparisons assessed here the evidence relating to key outcomes was assessed as being of low or very low certainty. In some cases this was the result of evidence being at high risk of bias, but in more instances it was a consequence of serious imprecision or inconsistency, or both; in some cases indirectness was an issue due to the use of surrogate outcomes. Although we judged a minority of studies to be at high risk of bias, we judged most to be at unclear risk of bias on several or most domains. Often the fact that there was only a single study available ‐ or only a single study with analysable data ‐ meant that confidence intervals were very wide or fragile, or both, because of the small number of participants represented. A number of studies adopted an intra‐individual design (see Potential biases in the review process) and it was unclear whether this was taken into consideration in the analyses. There is, therefore, a high level of uncertainty around many of the findings. We note that this is the case although we adopted a conservative approach to downgrading for risk of bias in our GRADE assessments, and only downgraded where there was judged to be a high risk of bias: we did not downgrade for risk of bias where one or more domains had an unclear risk of bias.

Potential biases in the review process

Study design

A number of studies adopted an intra‐individual (split‐body) approach analogous to the 'split‐mouth' design (Lesaffre 2009). These studies have particular issues and, if incorrectly analysed, can produce inaccurate confidence intervals around the estimates of effect. Where there are a number of such studies for a given comparison there is a case for analysing them separately from parallel‐group designs. We had failed to anticipate the number of trials with these designs, which were eligible for inclusion in our review, and therefore our approach to handling them is necessarily post‐hoc. There were ten trials with this design and it was unclear whether they had accounted for the intra‐individual design in their analyses. In most of our analyses there were limited numbers of these studies as they were distributed across the large number of comparisons in the review. Therefore we have adopted a pragmatic and conservative approach: where these studies contributed data to a meta‐analysis with at least two other studies, we conducted a post‐hoc sensitivity analysis and used the results of that to inform the GRADE assessment if it differed substantively from the primary analysis. Where there was only one additional study in the analysis, we reported both the pooled results and the results of the two trials with different designs separately. Where these studies were present in an analysis but did not contribute weight to it (because of zero events or lack of measures of variance), we noted their presence. In all except one case the sensitivity analysis conducted did not materially affect the estimate of effect or the confidence intervals. In a single case we have downgraded twice rather than once for imprecision because, in the sensitivity analysis excluding a trial with an intra‐individual design, the confidence intervals differed from the main analysis in crossing the line of no effect. We are therefore confident that our post‐hoc approach to data from these trials is unlikely to have affected the findings of the review, and that fully including the data increases the comprehensiveness of the review.

Language and setting

Eleven of the included studies were reported in languages other than English, with ten in Chinese and one in Portuguese, as were many of the excluded studies (Chinese, German). We therefore do not believe that language bias is likely to be an issue. The included studies were conducted across a wide range of countries. Only around a third (17) of the studies were conducted in Western, high‐income countries. The majority were based in low‐ to middle‐income countries, almost all in Asia, where much of the mortality and morbidity burden from burns is concentrated. It therefore seems likely that in this respect participants in the included studies may reflect those with burns world wide.

Funding

The great majority of the included studies did not state how they were funded. Of those where the funding source was clear, five were funded by industry and six by other non‐commercial sources; two others reported both types of funding. Where funding sources are not clearly reported, it can be conservative to assume that this may be a source of bias. However, in this case many of the studies were small and of short duration, and it is therefore likely that they may not have received any external funding.

Publication bias

We did not find evidence of publication bias although it remains a possibility that undetected publication bias was present in some analyses. In some comparisons it was clear that the antiseptic treatment was intended as the comparator: the intervention that the trial was designed to evaluate was the non‐antibacterial comparator. If either funding or selective publication were leading to the introduction of bias or potential bias, this would mean that trials that favoured antiseptics would be disproportionately likely to be absent. This was not a pattern that we found evidence to support.

Agreements and disagreements with other studies or reviews

There is a current published Cochrane Review of antibiotics for the prevention (prophylaxis) of burn wound infection (Barajas‐Nava 2013), while a second Cochrane Review of antibiotics for the treatment of infected burn wounds is now underway (Lu 2016). This review of antiseptics complements these reviews and completes the assessment of evidence for agents with antimicrobial properties in the care of all burn wounds, whether infected or not.

There is some overlap between this review and other Cochrane and non‐Cochrane reviews of dressings for partial‐thickness burns (Wasiak 2013), and of individual agents with antiseptic properties for all types of wounds (Aziz 2012; Dat 2012; Jull 2015; Storm‐Versloot 2010; Vermuelen 2010), however, this review provides a single synthesis of the randomised evidence relating to all antiseptics for any type of burn wound as well as having a more recent search. This, together with differences in inclusion criteria mean that there are differences in the included studies. It is worth noting that over 30% of the studies in this review were published in 2010 or later. There are also differences in the approach to analysis, with this review deriving hazard ratios to allow evaluation of the 'chance' of healing over time for some of the comparisons; this is a more robust measure of the outcome than mean time to healing or the occurrence of healing events at a single time point.

Authors' conclusions

Implications for practice

The effect of different treatments in many of the comparisons is unclear: it is often uncertain whether the antiseptics assessed in these (often single, small) trials are associated with any difference in healing, infections, adverse events or other outcomes. The certainty of this evidence is low or very low, primarily due to the high levels of imprecision around the estimates of effect.

In some cases (see Summary of main results) there is moderate or high evidence for the comparisons of honey to topical antibiotics or non‐antibacterial dressings. This suggested that there is an advantage to the use of honey over the alternative treatments in these comparisons in terms of wound healing. We note that there was very limited reporting of data on pain in the comparisons involving honey. Pain is particularly important in this patient group and has been reported to be a consideration in the use of honey. Practitioners may wish particularly to take the lack of data on this outcome into account, together with the evidence on healing and infection. There is, however, some moderate certainty evidence that pain may be reduced more from baseline in burns treated with Aloe Vera compared with silver sulfadiazine (SSD) and that there may be lower levels of pain in participants treated with cerium nitrate in addition to SSD compared with SSD alone.

Much of the evidence in this review will also need careful consideration by practitioners in order to determine whether it is relevant to their practice. There was a degree of heterogeneity in terms of the age of participants ‐ ranging from very young children to adults. However most of the studies ‐ with some notable exceptions ‐ focused on burns, which were described as, or corresponded to, 'second‐degree burns' and most were below 40% TBSA ‐ in some cases very much less than this. In addition some of the comparators used may not be considered by practitioners to be relevant to their clinical work. This is particularly the case for the comparisons involving honey and non‐antibacterial dressings. In many cases it is possible that the evidence may be only indirectly relevant to particular patient groups.

In many cases the methods used in the trials were not well described and we are unsure whether they were designed in a way that makes different types of bias unlikely; although we have not downgraded for an unclear risk of bias, we are not confident that it may not be present.

Implications for research

There is a surprising paucity of randomised evidence assessing comparisons between some of the principal antiseptic agents ‐ both with each other and with either topical antibiotics or non‐antibacterial agents. Many comparisons were represented by a single trial and many trials did not report adequate data on key outcomes. The exception to this is that there are a large number of trials that assess (1) silver‐based treatment (mainly dressings) compared with the topical antibiotic SSD and (2) honey compared with alternative treatments including SSD. Very few of these trials, however, are sufficiently clearly reported for us to be completely confident that they were well‐conducted. This is also the case for the smaller number of trials available for other comparisons. Most trials were also small, meaning that there is necessarily a high level of imprecision and often inconsistency present in the comparisons to which they contribute. Nine comparisons included only a single small trial. Where more than one trial contributed to the comparison, it was still sometimes the case that primary outcomes were reported by only one trial ‐ this was particularly the case for infection. In some comparisons there was a large difference in the results of trials reporting an outcome; for example in the comparison of wound healing for iodine compared with non‐antibacterial treatments. For all these reasons the evidence for most outcomes for most comparisons was assessed as being of low or very low certainty. Even where there was evidence that was assessed as moderate or high certainty, the reporting of the trials was often insufficient for us to be very confident that bias was unlikely.

Given the key importance of infection control as well as wound healing, the lack of evidence on this outcome for many comparisons was particularly striking. In view of this uncertainty and the large number of treatment options with antiseptic properties, the design of future trials should be driven by high priority questions from patients and other decision makers. It is also important for research to ensure that the outcomes that are collected in research studies are those that matter to patients and health professionals; clinical infection and pain may be examples of such outcomes. Where trials are conducted, good practice guidelines must be followed in their design, implementation and reporting. Such trials should be adequately powered to detect differences in time to healing, should use appropriate statistical methods for time‐to‐event analyses and should include adequate follow‐up to allow all participants to heal. Consideration should also be given to enrolment criteria to ensure that trials are relevant to patients with differing levels of burn severity (depth) and extent (proportion of total body surface area).

What's new

Acknowledgements

The authors would like to acknowledge the contribution to the protocol of peer referees David Margolis, Heather Cleland, Camila Pino, Christine Fyfe and Abimbola Farinde and the copy editor Clare Dooley; and the contributions to the review of peer referees Andrew Jull, Mark Rodgers, Camila Pino and Caitlin Mitchell. They are also grateful to the copy editor, Denise Mitchell.

We would also like to acknowledge the translation assistance of Mario Cruciani, Irina Telegina, Jennifer Brown, Rachel Riera, Ana Luiza C Martimbianco and Debra Fayter.

Appendices

Appendix 1. Search strategies

The Cochrane Central Register of Controlled Clinical Trials (CENTRAL)

#1 MeSH descriptor: [Anti‐Infective Agents] explode all trees
#2 MeSH descriptor: [Penicillins] explode all trees
#3 MeSH descriptor: [Cephalosporins] explode all trees
#4 MeSH descriptor: [Aminoglycosides] explode all trees
#5 MeSH descriptor: [Quinolones] explode all trees
#6 MeSH descriptor: [Clindamycin] explode all trees
#7 MeSH descriptor: [Metronidazole] explode all trees
#8 MeSH descriptor: [Trimethoprim] explode all trees
#9 MeSH descriptor: [Mupirocin] explode all trees
#10 MeSH descriptor: [Neomycin] explode all trees
#11 MeSH descriptor: [Fusidic Acid] explode all trees
#12 MeSH descriptor: [Framycetin] explode all trees
#13 MeSH descriptor: [Polymyxins] explode all trees
#14 MeSH descriptor: [Chlortetracycline] explode all trees
#15 (antibiotic* or antimicrobial* or antibacterial* or penicillin* or cephalosporin* or aminoglycoside* or quinolone* or clindamycin or metronidazole or trimethoprim or mupirocin or "pseudomonic acid" or neomycin or "fusidic acid" or framycetin or polymyxin* or chlortetracycline):ti,ab,kw
#16 MeSH descriptor: [Antisepsis] explode all trees
#17 antiseptic*:ti,ab,kw
#18 MeSH descriptor: [Soaps] explode all trees
#19 MeSH descriptor: [Iodophors] explode all trees
#20 MeSH descriptor: [Chlorhexidine] explode all trees
#21 MeSH descriptor: [Alcohols] explode all trees
#22 MeSH descriptor: [Hydrogen Peroxide] explode all trees
#23 MeSH descriptor: [Benzoyl Peroxide] explode all trees
#24 MeSH descriptor: [Gentian Violet] explode all trees
#25 MeSH descriptor: [Hypochlorous Acid] explode all trees
#26 MeSH descriptor: [Hexachlorophene] explode all trees
#27 MeSH descriptor: [Potassium Permanganate] explode all trees
#28 MeSH descriptor: [Silver] explode all trees
#29 MeSH descriptor: [Silver Sulfadiazine] explode all trees
#30 MeSH descriptor: [Honey] explode all trees
#31 ("soap" or "soaps" or iodophor* or povidone or iodine or chlorhexidine or betadine or "alcohol" or disinfectant* or "hydrogen peroxide" or "benzoyl peroxide" or "gentian violet" or hypochlorit* or eusol or dakin* or hexachlorophene or benzalkonium or "potassium permanganate" or "silver sulfadiazine" or "silver sulphadiazine" or honey*):ti,ab,kw
#32 #1 or #2 or #3 or #4 or #5 or #6 or #7 or #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27 or #28 or #29 or #30 or #31
#33 MeSH descriptor: [Burns] explode all trees
#34 ("burn" or "burns" or "burned" or scald*):ti,ab,kw
#35 ("thermal" near injur*):ti,ab,kw
#36 #33 or #34 or #35
#37 #32 and #36 in Trials

Ovid MEDLINE

1 exp Anti‐Infective Agents/
2 exp Penicillins/
3 exp Cephalosporins/
4 exp Aminoglycosides/
5 exp Quinolones/
6 exp Clindamycin/
7 exp Metronidazole/
8 exp Trimethoprim/
9 exp Mupirocin/
10 exp Neomycin/
11 exp Fusidic Acid/
12 exp Framycetin/
13 exp Polymyxins/
14 exp Chlortetracycline/
15 (antibiotic$ or antimicrobial$ or antibacterial$ or penicillin$ or cephalosporin$ or aminoglycoside$ or quinolone$ or clindamycin or metronidazole or trimethoprim or mupirocin or pseudomonic acid or neomycin or fusidic acid or framycetin or polymyxin$ or chlortetracycline).ti,ab.
16 exp Antisepsis/
17 antiseptic$.ti,ab.
18 exp Soaps/
19 exp Iodophors/
20 exp Chlorhexidine/
21 exp Alcohols/
22 exp Hydrogen Peroxide/
23 exp Benzoyl Peroxide/
24 exp Gentian Violet/
25 exp Hypochlorous Acid/
26 exp Hexachlorophene/
27 exp Potassium Permanganate/
28 exp Silver/
29 exp Silver Sulfadiazine/
30 exp Honey/
31 (soap$1 or iodophor$ or povidone or iodine or chlorhexidine or betadine or alcohol$1 or disinfectant$ or hydrogen peroxide or benzoyl peroxide or gentian violet or hypochlorit$ or eusol or dakin$ or hexachlorophene or benzalkonium or potassium permanganate or silver sulfadiazine or silver sulphadiazine or honey$).ti,ab.
32 or/1‐31
33 exp Burns/
34 (burn or burns or burned or scald*).tw.
35 (thermal adj injur*).tw.
36 or/33‐35
37 and/32,36
38 randomized controlled trial.pt.
39 controlled clinical trial.pt.
40 randomi?ed.ab.
41 placebo.ab.
42 clinical trials as topic.sh.
43 randomly.ab.
44 trial.ti.
45 or/38‐44 (1006117)
46 exp animals/ not humans.sh.
47 45 not 46
48 and/37,47

Ovid Embase

1 exp Antiinfective Agent/
2 exp Penicillin G/
3 exp Cephalosporin/
4 exp Aminoglycoside/
5 exp Quinolone/
6 exp Clindamycin/
7 exp Metronidazole/
8 exp Trimethoprim/
9 exp Pseudomonic Acid/
10 exp Neomycin/
11 exp Fusidic Acid/
12 exp Framycetin/
13 exp Polymyxin/
14 exp Chlortetracycline/
15 (antibiotic$ or antimicrobial$ or antibacterial$ or penicillin$ or cephalosporin$ or aminoglycoside$ or quinolone$ or clindamycin or metronidazole or trimethoprim or mupirocin or neomycin or fusidic acid or framycetin or polymyxin$ or chlortetracycline).ti,ab.
16 exp antisepsis/
17 antiseptic$.ti,ab.
18 exp Soap/
19 exp Iodophor/
20 exp Chlorhexidine/
21 exp Alcohol/
22 exp Hydrogen Peroxide/
23 exp Benzoyl Peroxide/
24 exp Gentian Violet/
25 exp Hypochlorous Acid/
26 exp Hexachlorophene/
27 exp Potassium Permanganate/
28 exp Silver/
29 exp Silver Sulfadiazine/
30 exp Honey/
31 (soap$1 or iodophor$ or povidone or iodine or chlorhexidine or betadine or alcohol$1 or disinfectant$ or hydrogen peroxide or benzoyl peroxide or gentian violet or hypochlorit$ or eusol or dakin$ or hexachlorophene or benzalkonium or potassium permanganate or silver sulfadiazine or silver sulphadiazine or honey$).ti,ab.
32 or/1‐31
33 exp burn/
34 (burn or burns or burned or scald*).tw.
35 (thermal adj injur*).tw.
36 or/33‐35
37 and/32,36
38 Randomized controlled trials/
39 Single‐Blind Method/
40 Double‐Blind Method/
41 Crossover Procedure/
42 (random* or factorial* or crossover* or cross over* or cross‐over* or placebo* or assign* or allocat* or volunteer*).ti,ab.
43 (doubl* adj blind*).ti,ab.
44 (singl* adj blind*).ti,ab.
45 or/38‐44
46 exp animals/ or exp invertebrate/ or animal experiment/ or animal model/ or animal tissue/ or animal cell/ or nonhuman/
47 human/ or human cell/
48 and/46‐47
49 46 not 48
50 45 not 49
51 and/37,50

EBSCO CINAHL

S45 S31 AND S44
S44 S32 OR S33 OR S34 OR S35 OR S36 OR S37 OR S38 OR S39 OR S40 OR S41 OR S42 OR S43
S43 TI allocat* random* or AB allocat* random*
S42 MH "Quantitative Studies"
S41 TI placebo* or AB placebo*
S40 MH "Placebos"
S39 TI random* allocat* or AB random* allocat*
S38 MH "Random Assignment"
S37 TI randomi?ed control* trial* or AB randomi?ed control* trial*
S36 AB ( singl* or doubl* or trebl* or tripl* ) and AB ( blind* or mask* )
S35 TI ( singl* or doubl* or trebl* or tripl* ) and TI ( blind* or mask* )
S34 TI clinic* N1 trial* or AB clinic* N1 trial*
S33 PT Clinical trial
S32 MH "Clinical Trials+"
S31 S26 AND S30
S30 S27 OR S28 OR S29
S29 TI thermal n1 injur* OR AB thermal n1 injur*
S28 TI ( burn or burns or burned or scald* ) OR AB ( burn or burns or burned or scald* )
S27 (MH "Burns+")
S26 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17 or S18 or S19 or S20 or S21 or S22 or S23 or S24 or S25
S25 TI ( soap* or iodophor* or povidone or iodine or chlorhexidine or betadine or alcohol* or disinfectant* or hydrogen peroxide or benzoyl peroxide or gentian violet or hypochlorit* or eusol or dakin* or hexachlorophene or benzalkonium or potassium permanganate or silver sulfadiazine or silver sulphadiazine or honey*) or AB ( soap* or iodophor* or povidone or iodine or chlorhexidine or betadine or alcohol* or disinfectant* or hydrogen peroxide or benzoyl peroxide or gentian violet or or hypochlorit* or eusol or dakin* or hexachlorophene or benzalkonium or potassium permanganate or silver sulfadiazine or silver sulphadiazine or honey*)
S24 (MH "Honey")
S23 (MH "Silver Sulfadiazine")
S22 (MH "Silver")
S21 (MH "Hexachlorophene")
S20 (MH "Gentian Violet")
S19 (MH "Hydrogen Peroxide")
S18 (MH "Alcohols+")
S17 (MH "Chlorhexidine")
S16 (MH "Povidone‐Iodine")
S15 (MH "Iodine")
S14 (MH "Soaps")
S13 TI antiseptic*
S12 S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11
S11 TI ( antibiotic* or antimicrobial* or antibacterial* or penicillin* or cephalosporin* or aminoglycoside* or quinolone* or clindamycin or metronidazole or trimethoprim or mupirocin or pseudomonic acid or neomycin or fusidic acid or framycetin or polymyxin* or chlortetracycline ) or AB ( antibiotic* or antimicrobial* or antibacterial* or penicillin* or cephalosporin* or aminoglycoside* or quinolone* or clindamycin or metronidazole or trimethoprim or mupirocin or pseudomonic acid or neomycin or fusidic acid or framycetin or polymyxin* or chlortetracycline)
S10 (MH "Polymyxins+")
S9 (MH "Neomycin")
S8 (MH "Mupirocin")
S7 (MH "Trimethoprim+")
S6 (MH "Metronidazole")
S5 (MH "Clindamycin")
S4 (MH "Aminoglycosides+")
S3 (MH "Cephalosporins+")
S2 (MH "Penicillins+")
S1 (MH "Antiinfective Agents+")

US National Institutes of Health Ongoing Trials Register (ClinicalTrials.gov)

[“antiseptic” OR “antibacterial”] AND “burn”

World Health Organization International Clinical Trials Registry Platform

[“antiseptic” OR “antibacterial”] AND “burn”

Appendix 2. Assessment of risk of bias

The Cochrane tool for assessing risk of bias

1. Was the allocation sequence randomly generated?

Low risk of bias

The investigators describe a random component in the sequence generation process such as: referring to a random number table; using a computer random‐number generator; coin tossing; shuffling cards or envelopes; throwing dice; drawing of lots.

High risk of bias

The investigators describe a non‐random component in the sequence generation process. Usually, the description would involve some systematic, non‐random approach, for example: sequence generated by odd or even date of birth; sequence generated by some rule based on date (or day) of admission; sequence generated by some rule based on hospital or clinic record number.

Unclear

Insufficient information about the sequence generation process provided to permit a judgement of low or high risk of bias.

2. Was the treatment allocation adequately concealed?

Low risk of bias

Participants and investigators enrolling participants could not foresee assignment because one of the following, or an equivalent method, was used to conceal allocation: central allocation (including telephone, web‐based and pharmacy‐controlled randomisation); sequentially‐numbered drug containers of identical appearance; sequentially‐numbered, opaque, sealed envelopes.

High risk of bias

Participants or investigators enrolling participants could possibly foresee assignments and thus introduce selection bias, such as allocation based on: use of an open random allocation schedule (e.g. a list of random numbers); assignment envelopes without appropriate safeguards (e.g. envelopes were unsealed, non‐opaque, or not sequentially numbered); alternation or rotation; date of birth; case record number; any other explicitly unconcealed procedure.

Unclear

Insufficient information provided to permit a judgement of low or high risk of bias. This is usually the case if the method of concealment is not described, or not described in sufficient detail to allow a definite judgement, for example if the use of assignment envelopes is described, but it remains unclear whether envelopes were sequentially numbered, opaque and sealed.

3. Blinding ‐ was knowledge of the allocated interventions adequately prevented during the study?

Low risk of bias

Any one of the following.

• No blinding, but the review authors judge that the outcome and the outcome measurement are not likely to be influenced by lack of blinding.

• Blinding of participants and key study personnel ensured, and unlikely that the blinding could have been broken.

• Either participants or some key study personnel were not blinded, but outcome assessment was blinded and the non‐blinding of others unlikely to introduce bias.

High risk of bias

Any one of the following.

• No blinding or incomplete blinding, and the outcome or outcome measurement is likely to be influenced by lack of blinding.

• Blinding of key study participants and personnel attempted, but likely that the blinding could have been broken.

• Either participants or some key study personnel were not blinded, and the non‐blinding of others likely to introduce bias.

Unclear

Either of the following.

• Insufficient information to permit judgement of low or high risk of bias.

• The study did not address this outcome.

4. Were incomplete outcome data adequately addressed?

Low risk of bias

Any one of the following.

• No missing outcome data.

• Reasons for missing outcome data are unlikely to be related to true outcome (for survival data, censoring unlikely to be introducing bias).

• Missing outcome data are balanced in numbers across intervention groups, with similar reasons for missing data across groups.

• For dichotomous outcome data, the proportion of missing outcomes compared with the observed event risk is not enough to have a clinically relevant impact on the intervention effect estimate.

• For continuous outcome data, a plausible effect size (difference in means or standardised difference in means) among missing outcomes is not enough to have a clinically relevant impact on the observed effect size.

• Missing data have been imputed using appropriate methods.

High risk of bias

Any one of the following.

• Reasons for missing outcome data are likely to be related to the true outcome, with either an imbalance in numbers or reasons for missing data across intervention groups.

• For dichotomous outcome data, the proportion of missing outcomes compared with the observed event risk is enough to induce clinically relevant bias in the intervention effect estimate.

• For continuous outcome data, a plausible effect size (difference in means or standardised difference in means) among missing outcomes is enough to induce a clinically relevant bias in the observed effect size.

• 'As‐treated' analysis done with a substantial departure of the intervention received from that assigned at randomisation.

• Potentially inappropriate application of simple imputation.

Unclear

Either of the following.

• Insufficient reporting of attrition/exclusions to permit a judgement of low or high risk of bias (e.g. number randomised not stated, no reasons for missing data provided).

• The study did not address this outcome.

5. Are reports of the study free of suggestion of selective outcome reporting?

Low risk of bias

Either of the following.

• The study protocol is available and all of the study’s prespecified (primary and secondary) outcomes that are of interest in the review have been reported in the prespecified way.

• The study protocol is not available but it is clear that the published reports include all expected outcomes, including those that were prespecified (convincing text of this nature may be uncommon).

High risk of bias

Any one of the following.

• Not all of the study’s prespecified primary outcomes have been reported.

• One or more primary outcomes is/are reported using measurements, analysis methods, or subsets of the data (e.g. subscales) that were not prespecified.

• One or more reported primary outcomes was/were not prespecified (unless clear justification for their reporting is provided, such as an unexpected adverse effect).

• One or more outcomes of interest in the review is/are reported incompletely so that they cannot be entered in a meta‐analysis.

• The study report fails to include results for a key outcome that would be expected to have been reported for such a study.

Unclear

Insufficient information provided to permit a judgement of low or high risk of bias. It is likely that the majority of studies will fall into this category.

6. Other sources of potential bias

Low risk of bias

The study appears to be free of other sources of bias.

High risk of bias

There is at least one important risk of bias. For example, the study:

• had a potential source of bias related to the specific study design used; or

• has been claimed to have been fraudulent; or

• had some other problem.

Unclear

There may be a risk of bias, but there is either:

• insufficient information to assess whether an important risk of bias exists; or

• insufficient rationale or evidence that an identified problem will introduce bias.

Appendix 3. Risk of bias in cluster‐randomised trials

In cluster‐randomised trials, particular biases to consider include:

• recruitment bias;

• baseline imbalance;

• loss of clusters;

• incorrect analysis; and

• comparability with individually‐randomised trials.

Recruitment bias: can occur when individuals are recruited to the trial after the clusters have been randomised, as the knowledge of whether each cluster is an 'intervention' or 'control' cluster could affect the types of participants recruited.

Baseline imbalance: cluster‐randomised trials often randomise all clusters at once, so lack of concealment of an allocation sequence should not usually be an issue. However, because small numbers of clusters are randomised, there is a possibility of chance baseline imbalance between the randomised groups, in terms of either the clusters or the individuals. Although this is not a form of bias as such, the risk of baseline differences can be reduced by using stratified or pair‐matched randomisation of clusters. Reporting of the baseline comparability of clusters, or statistical adjustment for baseline characteristics, can help reduce concern about the effects of baseline imbalance.

Loss of clusters: occasionally complete clusters are lost from a trial, and have to be omitted from the analysis. Just as for missing outcome data in individually‐randomised trials, this may lead to bias. In addition, missing outcomes for individuals within clusters may also lead to a risk of bias in cluster‐randomised trials.

Incorrect analysis: many cluster‐randomised trials are analysed by incorrect statistical methods that do not take the clustering into account. Such analyses create a 'unit of analysis error' and produce over‐precise results (the standard error of the estimated intervention effect is too small) and P values that are too small. They do not lead to biased estimates of effect. However, if they remain uncorrected, they will receive too much weight in a meta‐analysis.

Comparability with individually‐randomised trials: in a meta‐analysis that includes both cluster‐randomised and individually‐randomised trials, or includes cluster‐randomised trials with different types of clusters, possible differences between the intervention effects being estimated need to be considered. For example, in a vaccine trial of infectious diseases, a vaccine applied to all individuals in a community would be expected to be more effective than a vaccine applied to only half the people. Another example is provided by discussion of a Cochrane Review of hip protectors (Hahn 2005), where cluster trials showed a large positive effect, whereas individually‐randomised trials did not show any clear benefit. One possibility is that there was a 'herd effect' in the cluster‐randomised trials (which were often performed in nursing homes, where compliance with using the protectors may have been enhanced). In general, such 'contamination' would lead to underestimates of effect. Thus, if an intervention effect is still demonstrated despite contamination in those trials that were not cluster‐randomised, a confident conclusion about the presence of an effect can be drawn. However, the size of the effect is likely to be underestimated. Contamination and 'herd effects' may be different for different types of cluster.

Appendix 4. Extracted subgroup data for wound healing

Notes

Edited (no change to conclusions)

Data and analyses

Comparison 1

Silver dressings versus topical antibiotics

Analysis

Comparison 1 Silver dressings versus topical antibiotics, Outcome 12 Cost‐effectiveness/wound healed (21 days).

Comparison 2

Honey versus topical antibiotics

Analysis

Comparison 2 Honey versus topical antibiotics, Outcome 2 Wound healing (risk ratio) (up to 60 days).

Comparison 3

Aloe vera vs topical antibiotics

Comparison 4

Iodine‐based treatments versus topical antibiotics

Comparison 5

Silver‐based antiseptics versus non‐antimicrobial

Analysis

Comparison 5 Silver‐based antiseptics versus non‐antimicrobial, Outcome 2 Positive swab (21 days).

Comparison 6

Honey versus non‐antibacterial dressing

Comparison 7

Chlorhexadine versus non‐antibacterial dressing

Analysis

Comparison 7 Chlorhexadine versus non‐antibacterial dressing, Outcome 1 Wound healing (mean time to healing).

Comparison 8

Iodine‐based antiseptics versus non‐antibacterial treatments

Comparison 9

Cerium nitrate versus non antibacterial treatment

Characteristics of studies

Characteristics of included studies [ordered by study ID]