This review will follow the guidelines set out by the Collaboration for Environmental Evidence [12] and the ROSES reporting standards [13]. The ROSES form is available as Additional file 1.
Searching for articles
A structured search will be conducted in English using web-based search-engines (n = 1), organizational websites (n = 40), and platform databases (n = 1). A full list of all sources that will be used can be found in later in this section. An open call to stakeholders (loosely defined as anyone working in the fishing industry, as well as conservationists and researchers) through both formal and informal networks and over social media will also be asked to provide any sources they feel to be relevant to the review topic. Stakeholders may provide either study titles or provide links to relevant studies. Supplementary searches will be carried out based on stakeholder recommendations as well as through snowball sampling of key articles and studies (identified by stakeholders).
All efforts will be made to obtain full copies of articles identified in the search process. If an article is unobtainable through stakeholder networks, online databases, or web-based searches the authors (or their institutions) will be approached to provide a copy.
All search hits from the identified databases and specialist sources will be exported before being assessed using the inclusion criteria defined in “Eligibility criteria” section, but the hits from the identified search engines will be limited to the first 1000 results, as is standard in other systematic review protocols [14, 15].
The search terms used to find relevant data to answer the research question were chosen specifically for this review and were generated using the R package litsearchr, which partially automates keyword selection to generate a Boolean search string using a keyword co-occurrence network [16].
To generate the final Boolean Search string an initial naïve search (made up of broad search terms “impact static fish gear OR longline OR pot OR trap”) was conducted in Web of Science and Scopus. The results were imported into R. Any duplicates were removed, after which the package systematically extracted all potential keywords from the titles and abstracts. A keyword co-occurrence network was created, and potential keywords were grouped into themes echoing the PICO structure and were used to inform the final Boolean search string. A shorter search string was also created to be used in databases where the number of searchable words is capped. The results for this shorter search in each database will be compiled and any duplicates removed.
Browsing history will be deleted before conducting any database searches.
The following search string, generated through the litsearchr package, will be used where there are no character limits imposed by the database or search engine:
Fish* AND (creel* OR dam* OR demersal OR “drum net”* OR “eel tube” OR fence OR fixed OR “fixed gear” OR gillnet OR “gill net” OR gill-net OR longlin* OR long-lin* OR “long lin*” OR passive OR passive-* OR pot* OR static OR trammel OR trammel-* OR trap*) AND ((benth* OR bottom OR habitat+ OR seabed OR sea-bed OR “sea bed”) OR (affect* OR alter* OR caus* OR chang* OR damag* OR declin* OR decreas* OR deterior* OR effect* OR impact* OR increase* OR influenc* OR reduc* OR result* OR shift* OR transform*))
Shorter search terms will be used where character limits are in place or—as is the case for google scholar—where the database or search engine is limited in its ability to combine multiple Boolean values, for example:
Fish* AND (fixed OR net OR line OR passive OR pot* OR static OR trap*) AND (abundance OR “community composition” OR condition OR diversity OR mortality OR recovery rate OR richness)
Both the longer and shorter search strings were tested against a set of fifteen articles (see Additional file 2) which had been identified by subject experts and stakeholders as being highly relevant to answer the research questions. Both search strings were tested against these articles through searches in Google Scholar, Web of Science, Scopus, JSTOR, Oxford Academic, and the British Library’s e-thesis Online Service. Using a combination of both the long and short search strings all 15 test articles were retrieved (see Additional file 3).
The search engine Google Scholar (https://scholar.google.com/) and the platform of databases Web of Knowledge (https://apps.webofknowledge.com/WOS_GeneralSearch) will be searched.
Stakeholder consultation, combined with a web search of organizations involved in fisheries management and environmental conservation were used to identify the following organizational websites which will be searched for additional studies not available through bibliographic databases:
For each source the relevancy of first 50 hits will be checked (as is standard in other systematic review protocols), with all relevant papers, pages, or data being exported before being assessed using the previously defined inclusion criteria.
Other organizational sources may be consulted in addition to those listed as part of stakeholder feedback or through snowballing references from searches.
The first 1000 results returned by Google and Google Scholar will be downloaded, as is standard practice in other systematic review protocols [14, 15].
Search results will be imported and managed using the Endnote X9 reference management software. In cases where an article or paper cannot be imported into the software a separate file will be manually created. Once the search protocol has been completed reference files will be checked for duplicates and all duplicates will be removed.
Article screening and study eligibility criteria
Screening process
After duplicate articles are removed using Endnote studies will then be screened to ensure articles that do not provide relevant data are removed.
The screening criteria is based on the eligibility criteria provided in the following section.
Screening will occur in three stages: (1) screening the title according to its relevance, (2) screening the abstract according to its relevance, and (3) full text screening. At each stage, a random subset of at least 10% of will be subjected to another round of screening by a second independent reviewer to ensure accuracy and repeatability of the process. Where the level of agreement is below 0.6 according to a kappa test, all disagreements will be discussed in detail and further consistency checking will be conducted on an additional set of articles.
A list of articles that were removed at each stage of the screening process will be provided, alongside reasons for their exclusion.
Reviewers will not screen any studies they have authored.
Eligibility criteria
Eligibility criteria will be based on the PICO model. For a study or report to be considered eligible they must provide or signpost data on the following:
-
Population: Studies must perform experiments on marine benthic habitats, communities, or populations of macrofauna.
-
Exposure: Exposure to any type of static fishing gear.
-
Comparator: Control sites or areas with no intervention (i.e. no static fishing) will be used as a comparator (including before/after sites). However, control sites with very low levels of static fishing or even mobile fishing activity will also be included, as will fishing gradient studies that include a static fishing gear component.
-
Outcome: Outcomes include positive, negative, or no changes in the biodiversity of benthic fauna (measured by proxies such as diversity or species richness), changes in abundance (measured through % cover, density or biomass), changes in body size or size at maturity of an individual. Studies must include data on at least one benthic faunal species, specifying the name of either the taxon, genera, or species of the benthic fauna as well as values relating to numerical abundance, biomass, or diversity. Studies providing data on total faunal abundance will also be included.
-
Study design types: Experimental primary studies that use control and treatment areas, ‘before and after’, ‘control and impact’, or combinations of the two (BACI) will be considered. Comparisons of at least two sites experiencing different levels of bottom fishing exposure using static fishing gear will also be considered acceptable. Studies conducted in laboratory settings will also be included.
Study validity assessment
Critical appraisal will be done on a study-by-study basis. If an article reports more than one study each of these studies will undergo an individual critical appraisal. Studies meeting the eligibility criteria will be evaluated to gauge their internal and external validity and consequently classified as with having either a low, medium, or high potential for bias using the pre-defined framework CEE Critical Appraisal Tool (CEECAT) [17] which considers selection, performance, attrition, and reporting bias.
Data will be extracted from all studies, and a sensitivity analysis will be conducted to compare the outcomes between “low”, “moderate” and “high” risk papers. Results from the validity assessment will be recorded and presented with the results of the final review.
Validity assessment will be carried out independently by two reviewers. Any disagreements will be discussed, and a third reviewer will be consulted if no conclusion can be reached. Reviewers will not appraise any studies they have authored.
Data coding and extraction strategy
Data on study design, exposure, study results, habitat and geographical context will be extracted from included studies. Outcome data (such as sample sizes, means and measures of variation such as confidence intervals, standard deviations, and standard errors) will also be extracted. Information on effect modifiers will also be collected where available as well as metadata on study methodologies and general article identifiers. Summary statistics will be calculated if only raw data is provided. Where necessary, authors of the original study will be asked to provide unpublished primary data or provide clarification for unclear data. A list of data to be extracted can be found in Additional file 4.
Extraction will be carried out by one reviewer, however prior to full extraction of the data the extraction process will be independently tested. If the number of studies is > 50 then 10% of the studies will be tested by two reviewers. If the number of studies is < 50 then 30% of the studies will be tested. Any uncertainty will be discussed with the wider team until an agreement is reached. All extracted data will be saved in a Microsoft Excel spreadsheet which will be included in the final review as an additional file.
Potential effect modifiers/reasons for heterogeneity
All effect modifiers identified in the data extraction process will be recorded in an Excel spreadsheet. The following list identifies potential effect modifiers and was collated by the author team in consultation with stakeholders:
-
Benthic habitat type.
-
Depth of sampling site.
-
Environmental factors (e.g. fronts, reefs, mounds, nutrient cycling).
-
Fishing intensity.
-
Gear design (e.g. pot aperture size, J vs Circle hooks).
-
Gear type (i.e. pots, traps, longlines, or nets).
-
Geographical co-ordinates.
-
Hauling frequency.
-
Historical fishing pressure in the area.
-
Exposure duration and seasonality.
-
Number of pots/traps/hooks/nets.
-
Species biological traits (e.g. mobility; sessile vs mobile).
-
Study duration and seasonality.
-
Study sample size.
-
Taxonomic or functional groupings.
-
Time interval(s) between impact and sampling.
Additional effect modifiers that are identified during the review will be added to the list. All effect modifiers will be coded and included analysis.
Data synthesis and presentation
This study will employ both narrative and quantitative synthesis techniques. A narrative synthesis will be carried out on data from all studies and will describe the validity of results and tabulate study design, outcome measures, and other key descriptors. Maps of locations of all studies and of studies included in meta-analyses will be included. Heatmaps will be created by cross-tabulating different key descriptors to identify knowledge gaps.
Valid studies (as defined in “Eligibility criteria” section) which have comparable outcome effect sizes will be standardized and weighted appropriately. Where there is sufficient quantitative data, meta-analysis will be used to assess the effect of static fishing gear on benthic biodiversity. This will be conducted using the R package metafor which supports meta-regression analysis with both continuous and categorical moderators and fixed and random-effect models [18]. Meta-analysis will be conducted using both the estimates described by the authors as being the main results as well as all estimates from the same study. Effect sizes will be measured using the natural log-transformed response ratio [19].
The heterogeneity and the impact of effect modifiers will be explored through meta-regression analysis, which will help identify which effect modifiers have the greatest impact. Meta-regression will be conducted for all estimates.
Publication bias will be tested for using funnel plots, Egger tests, and comparisons of peer-reviewed and grey literature. A sensitivity analysis, comparing the outcomes of including and excluding papers with a “high risk” of bias, will be carried out to test the robustness of the validity assessment.
To avoid bias within the results reviewers carrying out this stage will not have published in this research area.