Types of studies

We included only randomised controlled trials (RCTs).

Types of participants

We included trials that included people of any age who had a burn wound covering more than 20% TBSA with more than 10% full‐thickness (i.e., a third‐degree burn). Trials involving any type of burn injury were eligible (e.g., chemical, scald or flame burns). We included trials with donor sites only when the donor sites were used to treat burn wounds.

Types of interventions

We considered studies for inclusion if recombinant human growth hormone (rhGH, any regimen) was compared with any comparable intervention, for example, testosterone analogues (e.g., oxandrolone), β‐blockade (propranolol) or placebo.

Types of outcome measures

Electronic searches

For this first update we searched the following electronic databases to find reports of relevant RCTs:

• Cochrane Wounds Group Specialised Register (searched 4 September 2014);

• The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2014, Issue 8);

• Database of Abstracts of Reviews of Effects (DARE) (The Cochrane Library 2014, Issue 3);

• Ovid MEDLINE (1950 to September Week 4 2014);

• Ovid MEDLINE (In‐Process & Other Non‐Indexed Citations September 8, 2014);

• Ovid EMBASE (1980 to 2014 Week 35);

• EBSCO CINAHL (1982 to 8 September 2014);

• PEDro, the Physiotherapy Evidence database (1980 to 8 September 2014);

• National Research Register (NRR) Archive (https://portal.nihr.ac.uk/Pages/NRRArchive.aspx);

• OAIster (http://oaister.umdl.umich.edu/o/oaister/) international institutional digital repository search engine;

• Web of Science (1975 to 8 September 2014);

• Dissertation abstracts (www.dissonline.de; www.theses.com; www.proquest.co.uk/products.pq/descriptions/pqdt.shtml), searched 8 September 2014;

• U.S. Food and Drug Administration (www.fda.gov), searched 8 September 2014;

• European Medicines Agency (www.ema.europa.eu), searched 8 September 2014.

We used the following strategy to search The Cochrane Central Register of Controlled Trials (CENTRAL):

#1 MeSH descriptor: [Burns] explode all trees 1195
#2 (burn or burns or burned or scald*):ti,ab,kw 3823
#3 (thermal* NEXT injur*):ti,ab,kw 154
#4 #1 OR #2 OR #3 4024
#5 MeSH descriptor: [Growth Hormone] explode all trees 3004
#6 (growth NEXT hormone*):ti,ab,kw 4425
#7 rhGH:ti,ab,kw 382
#8 somatotropin:ti,ab,kw 113
#9 #5 OR #6 OR #7 OR #8 4455
#10 #4 AND #9 65

Additional search strategies for Ovid MEDLINE, Ovid EMBASE and EBSCO CINAHL can be found in Appendix 2.

We did not restrict studies with respect to language, date of publication or study setting. We considered and included abstracts and letters if we were able to obtain complete manuscripts from the study author(s).

Searching other resources

Selection of studies

Two review authors independently screened the titles and abstracts of all records retrieved by the searches against the inclusion criteria to exclude obviously irrelevant studies. We retrieved as full text all citations that appeared to be relevant, and two review authors independently assessed these records against the inclusion criteria. In all instances, we resolved differences of opinion by discussion among the review authors.

Data extraction and management

Two review authors independently extracted data from the studies using paper data collection forms developed for this purpose. The standardised forms allowed the extraction of specific data, such as:

• country of origin;

• depth of the burn wound, size of the burn wound, burn type;

• care setting;

• number of participants randomised to each trial arm;

• eligibility criteria and key baseline participant data;

• details of the intervention (e.g., dosage, duration, type) that each group received;

• details of any co‐interventions;

• primary and secondary outcome(s) (with definitions);

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

• duration of follow‐up;

• number of withdrawals (by group);

• adverse events;

• funding sources.

We resolved all differences by discussion among the review authors. Each review author checked the data independently for accuracy and one review author (WET) entered the data into Review Manager 5 (RevMan 2011). The second review author verified the data entered (RSB).

Assessment of risk of bias in included studies

Two review authors independently assessed each included study using the Cochrane Collaboration tool for assessing the risk of bias (Higgins 2011a). This tool addresses six specific domains, namely: sequence generation, allocation concealment, blinding, incomplete outcome data, selective outcome reporting and other issues (e.g., extreme baseline imbalance) (Appendix 3). We assessed blinding and the completeness of outcome data for each outcome separately. We completed a 'Risk of bias' table for each eligible study and discussed any disagreements amongst all review authors to achieve a consensus. We present the assessment of the risk of bias using a 'Risk of bias' summary figure, which presents all of the judgements in a study‐by‐entry cross‐tabulation (Figure 1). This display of internal validity indicates the weight the reader may give to each study's results.

Figure 1

---

---

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

Measures of treatment effect

For data analysis, we followed the guidelines outlined in Section 9 of the Cochrane Handbook for Systematic Reviews of Interventions 5.0.1 (Deeks 2011).
We reported dichotomous outcomes (e.g., mortality) as risk ratios (RRs) with 95% confidence intervals (CIs). For continuous data, we presented results as the mean differences (MD) with 95% Cls. We used the standardised mean difference (SMD) when the studies assessed the same outcome, but used different scales (e.g., Vancouver Scar Scale and POSAS for scar assessment). The time to complete wound healing is time‐to‐event data and the most appropriate way of summarising this type of data is to use survival analysis methods and express the intervention effect as a hazard ratio. It is not appropriate to analyse time‐to‐event data using methods for continuous outcomes (e.g., using mean times‐to‐event) because the relevant times are only known for the subset of participants who had the event. When time is analysed as a continuous variable and not all participants experience the event (i.e., they are censored) they are excluded from the analysis which almost certainly introduces bias. When all of the participants in a study had completely healed wounds (the event), the mean wound‐healing time could be used.

We used Review Manager Version 5 software to generate the figures and statistical analyses (RevMan 2011).

Unit of analysis issues

The analysis of the included studies took into account the level at which randomisation occurred.

Dealing with missing data

We analysed data according to the intention‐to‐treat principle, i.e., all people included in the study at the point of randomisation were analysed according to their assigned treatment group, regardless of whether treatment was completed. Whenever possible, we contacted the original investigators to request missing data. We did not impute data if the missing data were not provided after several attempts to contact the study author. The potential impact of missing data on the review findings is addressed in the Discussion section.

Assessment of heterogeneity

The study synthesis methods depended on study quality, design and heterogeneity. We explored both clinical and statistical heterogeneity. In statistical heterogeneity the reported effect size is not consistent across studies. Clinical heterogeneity is characterised by a difference in study design; for instance, in the mixes of participants and in the implementation of interventions. We assessed heterogeneity between study results statistically using the I² statistic (Higgins 2003). This approach examines the percentage of the total variation across studies that is due to heterogeneity rather than chance. If the heterogeneity was more than 50% we used a random‐effects model instead of a fixed‐effect model when pooling was appropriate.

Assessment of reporting biases

We planned to assess reporting bias graphically using a funnel plot. We planned to report separately for each primary outcome; however, it must be recognised that if the number of eligible studies is small, the results are likely to be inconclusive (Egger 1997). The number of studies that could be pooled was very small (two to five studies), therefore this portion of the analysis was not undertaken.

Data synthesis

We examined clinical and methodological heterogeneity with reference to the study population (gender, age and TBSA percentage), intervention and outcome. In the absence of clinical and statistical heterogeneity, we used a fixed‐effect model to pool data. We investigated the cause of heterogeneity. In the presence of statistical heterogeneity (as estimated by an I² greater than 50%), we either applied a random‐effects model to pool the data or did not pool the data but undertook a narrative overview.

Subgroup analysis and investigation of heterogeneity

When we identified studies that included both adults and children, we presented the main effects for children and adults separately, data permitting.

Sensitivity analysis

We considered studies to have a high risk of bias if randomisation sequence generation, allocation concealment and outcome assessment blinding methods were inadequate or unclear. We performed a sensitivity analysis that omitted studies with a high risk of bias.