The target population of the benefit assessment was patients with wounds healing by primary intention. The experimental intervention was treatment with negative pressure wound therapy (NPWT). The comparator intervention was standard wound therapy (SWT).

The investigation considered the following patient-relevant outcomes:

• Mortality
• Wound closure
• Adverse events: wound complications and treatment complications (AEs)
• Amputation
• Pain
• Length of hospital stay and (re-)hospitalization
• Health-related quality of life
• Functioning
• Need of third-party help or need of long-term care

The outcomes “change in wound area or volume” as well as “change in wound surface after skin transplantation” were surveyed to provide supplementary information. Additionally, intervention-related and illness-related cost and patient satisfaction with treatment were to be considered, and related effects presented as supplementary information. Patient satisfaction was to be included in the analysis only to the extent it reflected health-related aspects.

Subjective outcomes (e.g., health-related quality of life) were considered only if they were surveyed using valid measuring instruments (e.g., validated scales).

Only randomized controlled trials (RCTs) were included in the benefit assessment. There were no restrictions regarding the study duration.

This benefit assessment is based on the results of the information retrieved by previous projects N04-03 and N06-02. The information retrieval was further updated for this report to include the period not covered by the searches for commissions N04-03 and N06-02 (2006 and later). The information was retrieved jointly for the benefit assessment of NPWT in patients with wounds healing by primary intention (N17-01B) and by secondary intention (N17-01A). After the respective study pools were defined, data were further processed in 2 separate benefit assessments.

A systematic search for primary literature was conducted in the databases MEDLINE, Embase and Cochrane Central Register of Controlled Trials. In parallel, a search for relevant systematic overviews was conducted in the databases MEDLINE, Embase, Cochrane Database of Systematic Reviews, and the HTA Database. Relevant systematic reviews had to be published in 2013 or later.

The following sources of information and search techniques were additionally used: trial registries, manufacturer documents, documents sent by the G-BA, reviews of reference lists, and documents made available from hearing procedures and author queries.

Relevant studies were selected by 2 reviewers independently from one another. Any discrepancies were resolved by discussion between the two reviewers. Data were extracted into standardized tables. To assess the qualitative certainty of conclusions, the risks of bias on both the study level and the outcome level was assessed and rated as high or low. The results of the individual studies were described broken down by outcomes.

Potential publication bias was suspected if the systematic search identified relevant completed studies without published results. To determine potential publication bias, the percentage of missing data (data gap) was calculated at study level. Potential outcome reporting bias was therefore disregarded. Studies with planned outcomes to be used exclusively for supplementary information were not taken into account since they were irrelevant for the conclusion. Studies without reported results which, according to the trial registry entry, were completed, prematurely terminated, or of unclear status were included in the calculations using their planned sample size, unless information to the contrary was available. This was done only if, at the time of the search, they should have been completed for more than 12 months and no usable data were supplied upon an author query.

The potential publication bias was assumed to have little effect on results if the patients from studies which had been completed for more than 12 months at the time of the search and for which no usable data were made available, even upon an author query, made up less than 10% of the total number of patients in the study pool. In this case, a regular benefit assessment was conducted since the missing data were not expected to have a relevant influence on results. If these patients represented between 10% and 30% of all patients, the potential publication bias was assumed to have a major effect on results. Since the missing data were expected to have a relevant influence on results, the certainty of conclusions from the benefit assessment was downgraded (proof to indication, indication to hint, hint to no hint). Due to the potential publication bias, the planned subgroup analyses were omitted. The surrogate validation of the outcomes “change in wound area or volume” and “change in wound surface” was not applicable for wounds healing by primary intention.

If the data gap involved more than 30% of patients, it was assumed that, due to potential publication bias, no conclusion could be drawn regarding benefit or harm, and no conclusion on benefit was inferred.

To categorize the completeness of data submissions by manufacturers, the percentage of missing data was calculated analogously to the procedure developed for potential publication bias. This was based on an agreement on the transfer and publication of study data entered into by the Institute and each involved manufacturer before data were submitted. This agreement applies to all projects, i.e. regardless of wound type distribution. In cases where a company was responsible for a relevant percentage of missing data at any tier, the selectively supplied data on patient-relevant outcomes were excluded.

For the sake of robustness testing, a second scenario was considered in each calculation of the percentage of missing data. For studies which reported no results and, according to the trial registry entry, were prematurely terminated or whose status is unclear, only half of the planned sample size was used in this scenario, unless information to the contrary was available.

Since mean differences can be influenced to varying degrees by wounds of different aetiologies being included in the studies, the latter were standardised in the meta-analyses using Hedges’ g whenever necessary.