This systematic review protocol has been developed according to the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines and the Cochrane Checklist of items to consider in data collection or data extraction (additional File 1). The review protocol has been registered a priori at the International Prospective Register of Systematic Reviews (PROSPERO ID: 142518).
The review will include mammalian animal subjects (e.g., rats and mice) and non-clinical human participants. Studies that use pre-clinical human participants (e.g., preexisting and undiagnosed fear of spiders or heights) or clinical samples (i.e., patients with mental health diagnoses) will not be included as these studies do not include a fear conditioning procedure and employ exposure therapy as the intervention.
The review will include experiments investigating extinction training in multiple spatial contexts. There are no restrictions to the timing, extent, duration, reinforcement schedule, use of a control stimulus (e.g., CS-), combination of interventions (e.g., multiple spatial contexts and massive extinction; see 22), and administration of nonreinforced trials within each set of extinction training. Critically, the intervention must aim to reduce conditioned fear in animals or humans. Studies that do not conduct fear conditioning prior to extinction will not be considered.
All included studies must have at least one control group that received comparable extinction treatment in one spatial context. These studies must also have outcome measures similar to their respective comparison groups. Furthermore, comparison groups should comprise of samples within the same population. Any study that does not utilise such procedures in the control group will be excluded.
All studies must contain at least one measurement of fear recovery by testing either spontaneous recovery, renewal (ABA or ABC renewal), or reinstatement after extinction training. For studies with non-clinical human samples, dependent variables include skin-conductance responses, heartrate, conditioned suppression, fear-potentiated startle responses, and US expectancy. Outlying dependent variables not listed above will be considered and fully discussed. For animal studies, conditioned suppression and suppression ratio will be reported. Data from studies that conduct progression of fear recovery (e.g., fear renewal followed by spontaneous recovery) will be included and statistical methods (e.g., three-level meta-analysis) will be used to account for the dependence of effect sizes within groups. A summary of findings table will be generated to highlight the main outcomes for informing future translational research into exposure therapy conducted in clinical settings.
The review will include laboratory-based primary research empirical studies. Subjects and participants must be randomised. All studies must, in a sequential order, contain a fear conditioning procedure, followed by manipulating extinction training with a direct impact on the conditioned response conducted across two or more spatial contexts, and test of recovery of the conditioned response. Fear conditioning is defined as pairing a neutral stimulus (e.g., tone) with a behaviorally relevant fearful stimulus (e.g., mild electric shock). Extinction across multiple spatial contexts is defined as receiving a set of extinction training in one environment, followed sequentially by receiving another set of extinction training in a different environment. An environment can be defined as a physical (e.g., 19), digital (e.g., 23), or virtual (e.g., 24) space, housed within the same locale (e.g., 10) or across different locations (e.g., 25), and can differ in dimensions, appearance (e.g., colour, lighting, patterns, and texture), and odour.
Language of publication
There will be no limitations on publication date, country of origin, or language. However, only search terms in the English language will be used (e.g., Spanish search terms will not be used). For articles published in other languages, language translation assistance will be sought within James Cook University Singapore and the National University of Singapore. If no translators are available, Google Translate® will be used to convert the article to English.
Studies that did not conduct fear conditioning prior to extinction will not be considered. Studies that did not conduct experiments (e.g., reviews) will not be included. Studies that investigate extinction across multiple temporal contexts (e.g., 26), extinction using multiple target and partner cues (e.g., 27), extinction-based exposure therapy (e.g., 28), cue or outcome interference (e.g., 29), reconsolidation blockade via drugs (e.g., 30) or reconsolidation blockade via extinction (e.g., 7) will not be included in this review. However, studies that use a combination of techniques, such as reconsolidation blockade and extinction in multiple spatial contexts, will be considered. Clinical studies will not be included due to the complexities of exposure therapy which embeds learning mechanisms such as psychoeducation and relaxation.
Information sources and search strategy
A comprehensive search with no restrictions on language will be conducted on PubMed, PsycINFO, Web of Science and Scopus for peer-reviewed studies on extinction in multiple contexts in animals and humans with the phrase:
((((multiple context extinction) OR (multiple contexts extinction)) OR (extinction in multiple contexts)) OR (extinction across multiple contexts)) OR (multiple AND contexts AND extinction). Search details such as date, location, and method will be recorded, and search results will be illustrated in a PRISMA flow diagram. Hand-searching will be done through forward and backward searching of references in eligible studies. Authors will be contacted to request for unpublished data.
Two independent reviewers will conduct the study selection. Studies identified through the search strategy will be extracted into a reference management software, EndNote (https://endnote.com/). The software will be used to filter duplicate studies which will then be manually checked by one reviewer to ensure any remaining duplicates are removed and quarantined.
This will consist of a 2-step title and abstract screening and full-text review process. The first screening process will begin with both reviewers independently conducting a screening of the titles and abstracts. This will ensure that each article gets screened by two reviewers. The list of identified articles for full-text screening will be contrasted and deliberated between the two reviewers. Following this, both reviewers will independently conduct full-text screening of a split of selected articles to determine inclusion. In addition, hand-searching will be done through forward and backward searching of references in individual studies. Both reviewers will need come to a consensus on the final list of included articles.
Data collection process
One reviewer will independently extract data from the list of identified articles. The data extracted will be reviewed by the second reviewer and discrepancies will be discussed before reaching a consensus. Data extraction will be piloted on a small number of identified studies. Information will be extracted onto a Microsoft Excel® sheet and include general article information (first author, year, title); aim of experiment (experiment number, objectives, outcomes, type of experiment); study population (species, type of sample, total sample, number of groups, sample size per group, mean age); study design (experimental paradigm, recovery mechanism, types of stimuli, types of contexts, counterbalancing strategies, context equalisation, context exposure, context preexposure, number of acquisition, extinction, and test contexts, trial duration, intertrial interval, interstimulus interval, retention interval); outcome measurement (conditioned suppression, conditioned suppression ratio, skin-conductance response, heartrate, US expectancy). Authors will be contacted via email to request for data clarifications or unpublished data. Requests to authors will be followed-up twice and the article will be rejected if the author does not reply within 1 month of the initial email.
Assessment of risk of bias in included studies
The full range of studies involving manipulation of multiple extinction contexts involves animals and humans. Two tools will be used for evaluation of the risk of bias. The Cochrane Risk of Bias tool version 2 (RoB 2) will be used to evaluate risk of bias in all human studies (31). RoB 2 assesses bias on five domains – randomisation process, deviation from intended intervention, missing outcome data, measurement of outcome and selection of the reported result. The Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE) will be used as the RoB tool for all animal studies included in the review. The SYRCLE assesses for potential issues with selection bias, performance bias, detection bias, attrition bias and reporting bias (32). Each study will be evaluated by the appropriate RoB tool by the first reviewer and evaluated by the second reviewer. Discrepancies will be discussed before reaching a consensus. The quality of the evidence will be assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE).
A narrative synthesis of the findings will be provided. This will include the type of measure used to determine recovery from extinction, the conceptualisation of contexts and study-specific manipulation of different contexts, the type of conditioned response undergoing extinction, and species.
It is anticipated that the variation of data analysis would vary widely across studies. Hence, following precedent literature (15), effect sizes will be based on 1) within-group differences in fear between the last extinction trial to the first test; 2) after between-group differences in fear have been shown to be nonsignificant, followed by observing recovery at test. In the event where a study employs both types of analyses, the first method will be used. For binary outcomes, a standard estimation of risk ratio (RR) and its 95% confidence interval (CI) will be used, and studies that use odds ratios (OR) will be transformed to RRs. The RR summary statistic is chosen as it is deemed more intuitive than ORs (33), and ORs are regarded as relative risks across clinical trials (34). For continuous outcomes, means, standard deviations, F tests, p- and t-values will be used. t-tests will be assumed to be two-sided unless specifically reported to be one-sided. Significance values will be assumed to be p = .05 unless specifically reported otherwise. Analyses will be conducted using the Review Manager (RevMan 5) tool. To account for the small sample sizes across studies, Hedges’ g will be used as a measure of effect size (35), where small, moderate, and large effects fall within 0.2, 0.5 and 0.8 respectively (36). A three-level meta-analysis will be conducted to account for studies that conduct progression of fear recovery (e.g., fear renewal followed by spontaneous recovery) that result in multiple effect sizes from the same sample (37–40; e.g., 15). Aggregate effect sizes will be calculated separately for animal fear conditioning studies, and non-clinical human fear conditioning studies. In addition, aggregate effect sizes will be calculated for each and all (combined) recovery outcomes (i.e., renewal, spontaneous recovery, reinstatement) to ascertain the efficacy of multiple spatial contexts extinction for each recovery type.
Assessment of statistical heterogeneity
Heterogeneity will be evaluated using the Q statistic and I2 statistic: the Q statistic detects the presence of heterogeneity where a significant Q statistic value suggests a presence of heterogeneity, while the I2 statistic calculates the extent of heterogeneity whereby I2 values of 25%, 50% and 75% indicate a low, moderate, and high heterogeneity respectively. Further prespecified subgroup analyses (see Table 1) will be conducted if moderate heterogeneity is detected to assess the influence of moderators.
List of potential subgroups
Context Conditioning Design
Humans / Animals
ABA / ABC
Following the recommendation of the Cochrane handbook for systematic reviews of interventions, publication bias will be assessed if an outcome contains at least 10 studies (41). Eggers’ test of intercept will be used to assess the existence of asymmetry, and therefore publication bias, in a funnel plot (42).