Types of studies

Randomised controlled trials (RCT), including cluster‐randomised trials.

Types of participants

Adults aged 18 years and older, irrespective of employment or health status, who have been exposed to stress or trauma in the past, or who are facing substantial stress or trauma currently or will be in the future (see Data synthesis).

We will include studies involving participants less than 18 years of age, as well as those aged 18 years and older, if data for participants aged 18 years and above are reported separately or can be obtained by contacting the study authors.

Types of interventions

Any psychological resilience intervention, irrespective of content, duration, setting or delivery model.

For the purpose of this review, psychological resilience interventions are defined as follows: interventions focused on fostering resilience or the related concepts of hardiness or post‐traumatic growth by strengthening well‐evidenced resilience factors that are thought to be modifiable by training (see above and Appendix 1; level 1).

We will only consider studies on pharmacological (e.g. treatment with antidepressants) and physical (e.g. exercise) interventions, as well as relaxation techniques (e.g. progressive muscle relaxation), if these interventions are part of psychological resilience training.

We will not include studies that merely examine the efficacy of disorder‐specific psychotherapy (e.g. cognitive behavioural therapy for depression). We will include broader, health‐promoting interventions (e.g. well‐being therapy) providing they focus on resilience and they address any of the resilience factors described above.

Types of outcome measures

Due to the different ways in which resilience has been operationalised, as well as the possible inclusion of broader, health‐promoting interventions, resilience as an intervention outcome cannot always be guaranteed in trials. Therefore, we will also define assessments on psychological adaptation (e.g. mental health) as primary outcomes.

Secondary outcomes include a range of psychological factors associated with resilience according to the current state of knowledge that are selected based on conceptual clarity and measurability (level 1a and 1b; see Appendix 1). We may include additional secondary outcomes after the full literature review has been conducted.

Measures for the assessment of psychological resilience and psychological adaptation as well as resilience factors are specified on the basis of previous reviews on resilience interventions (Leppin 2014; Macedo 2014; Robertson 2015; Vanhove 2015) and reviews on resilience measurements (Pangallo 2015; Windle 2011b) (see Appendix 3, Appendix 4, Appendix 5, respectively). We will examine the influence of the differing underlying concept in resilience scales in a sensitivity analysis (intervention outcome versus personality characteristic) (see Sensitivity analysis).

We will consider self‐rated and observer‐ or clinician‐rated measures as well as study outcomes at all time frames. The missing reporting of the above described primary or secondary outcomes is not an exclusion criterion in this review.

Electronic searches

We will retrieve relevant trials from the electronic sources listed below.

• Cochrane Central Register of Controlled Trials (CENTRAL; current issue) in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register.

• MEDLINE Ovid (1946 to present).

• Embase Ovid (1980 to present).

• PsycINFO EBSCOhost (1840 to present).

• CINAHL EBSCOhost (Cumulative Index to Nursing and Allied Health Literature; 1981 to present).

• PSYNDEX EBSCOhost (1977 to present).

• Science Citation Index Web of Science (SCI; 1970 to present).

• Social Science Citation Index Web of Science (SSCI; 1970 to present).

• Conference Proceedings Citation Index ‐ Social Science & Humanities Web of Science (CPCI‐SSH; 1990 to present).

• Conference Proceedings Citation Index ‐ Science Web of Science (CPCI‐S; 1990 to present).

• International Bibliography of the Social Sciences ProQuest (IBSS; 1951 to present).

• Applied Social Sciences Index & Abstracts ProQuest (ASSIA; 1987 to present).

• ProQuest Dissertations & Theses (PQDT; 1743 to present).

• Cochrane Database of Systematic Reviews (CDSR; current issue) in the Cochrane Library.

• Database of Abstracts of Reviews of Effects (DARE; current issue) in the Cochrane Library.

• Epistemonikos (epistemonikos.org; all available years).

• ERIC EBSCOhost (Education Resources Information Center Institute of Education Sciences; 1966 to present).

• Current Controlled Trials (controlled‐trials.com; all available years).

• ClinicalTrials.gov (clinicaltrials.gov; all available years).

• World Health Organization International Clinical Trials Registry Platform (WHO ICTRP; who.int/trialsearch; all available years).

The search strategy for MEDLINE is presented in Appendix 6, and we will adapt the search terms and syntax for other databases. We will not restrict the searches to language, publication status or publication format. We will limit our search to the period 1 January 1990 to present, to account for the fact that the resilience concept and its operationalisation have developed significantly over the past decades (Fletcher 2013; Hu 2015; Kalisch 2015; Pangallo 2015). Because of the lack of homogeneity for the period 1990 to 2014 (Robertson 2015), it is likely that using a broader time frame would make it even more difficult to detect resilience‐training studies with similar resilience concepts and assessments. Moreover, it appears plausible to concentrate on the period 1990 to present since the idea of resilience as an outcome and modifiable process only emerged in recent years, and paved the way for the development of resilience‐promoting interventions (Bengel 2009; Southwick 2011). Therefore, the idea of promoting resilience by specific training is relatively new (Leppin 2014), which can also be seen in the review of Macedo 2014, who searched for studies on resilience‐enhancing interventions every year until 2013 but only found RCTs published after 1990.

As resilience‐training programmes should be adapted to scientific findings on a regular basis, and with the current research focusing on the detection of general resilience mechanisms (Kalisch 2015; Luthar 2000), the last two years will be especially important in synthesising the evidence on newly developed resilience training.

Searching other resources

In addition to the electronic search, we will inspect the reference lists of all identified RCTs and reviews, and contact researchers in the field as well as the authors of selected trials to check if there are any unpublished or ongoing studies. If data are missing or unclear, we will contact the respective author. We will also search for grey literature (for example, conference proceedings) in appropriate databases (see Electronic searches).

Selection of studies

Two review authors (AK, IH) will independently screen titles and abstracts in order to determine eligible studies. Clearly irrelevant papers will be excluded immediately. At full‐text level, eligibility will be also inspected in duplicate by the same two review authors (AK, IH) working independently. We will calculate inter‐rater reliability at both stages of screening (title and abstract screening and full‐text screening). We will record our decisions in a PRISMA flow diagram (Moher 2009).

We will assess the feasibility of the selection criteria a priori by screening a small number (50) of studies in order to attain acceptable inter‐rater reliability. In case screening results in poor feasibility of the eligibility criteria, we will revise the criteria based on a mutual team discussion.

Data extraction and management

We will develop a data extraction sheet (Appendix 7), based on Cochrane guidelines (Higgins 2011c), and test it on 10 randomly‐selected included studies. If the initial test of the data extraction sheet fails (e.g. insufficient agreement between review authors AK and IH), we will adapt the extraction sheet on the basis of a mutual team discussion. Review authors AK and IH will independently extract the data in duplicate. The extraction sheet will contain the following aspects: source and eligibility, study methods (e.g. design), allocation process, participant characteristics, interventions and comparators, outcomes and assessment instruments (means and standard deviations in any standardised scale), results and miscellaneous aspects. Both review authors will resolve any disagreements in data collection by discussion; where they cannot reach a consensus, a third review author (AC or KL) will arbitrate. If necessary, we will contact the study authors to seek additional information.

Assessment of risk of bias in included studies

Two review authors (AK, IH) will independently assess the risk of bias of the included studies. We will check the risk of bias for each trial using the criteria presented in the Cochrane Handbook for Systematic Reviews of Interventions, hereafter referred to as the CochraneHandbook (Higgins 2011d) (see Appendix 8). Any disagreements will be resolved by discussion or by consulting a third review author. In accordance with Cochrane’s tool for assessing risk of bias (Higgins 2011b), we will critically assess the following domains: sequence generation and allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), and selective outcome reporting (reporting bias). In the first part of the assessment, we will describe what was reported to have happened in the study for each domain before assigning a judgment regarding the risk of bias (low, high or unclear) for that entry.

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

If there are missing data within the RCTs, we will contact the original researchers to provide the missing information (e.g. outcome data). We will compute missing standard deviations of continuous outcomes on the basis of other statistical information (e.g. CIs, standard errors, t values, P values, F values) (Higgins 2011e).

If standard deviations can neither be recovered from reported results nor obtained from the authors, we will consider single imputation by the means of pooled within‐treatment standard deviations from all other studies, providing less than five studies have missing standard deviations. If more than five studies have missing standard deviations, we will perform multiple imputation on the basis of the hierarchical model fitted to the non‐missing standard deviations. We expect to find enough information in all papers to restore standard deviations from the reported results.

We will record missing data and attrition levels for each included trial in the ‘Risk of bias’ tables (beneath the 'Characteristics of included studies' tables). Moreover, we will conduct a sensitivity analysis to examine the consequences of excluding trials with high levels of missing data on the conclusions of the review (see Sensitivity analysis).

Assessment of heterogeneity

We will assess the presence of clinical heterogeneity by comparing the trial and study population characteristics across all eligible trials (e.g. by generating descriptive statistics). In accordance with the Cochrane Handbook (Deeks 2011), we will explore if studies are sufficiently homogenous in terms of participant characteristics, interventions and outcomes.

We will assess methodological diversity by inspecting included studies for variability in study design and risk of bias. In accordance with previous reviews, which have already described the great heterogeneity in resilience intervention studies (Leppin 2014; Macedo 2014; Robertson 2015; Vanhove 2015), we will also discuss different forms of diversity in full in our review.

To assess statistical heterogeneity between included trials within each pair‐wise meta‐analysis (i.e. heterogeneity in observed treatment effects that exceeds sampling error alone), we will rely on forest plots, Chi² test, tau² statistic and I² statistic, as suggested by Deeks 2011. In addition, we will consider G², to take small‐study effects into account (Rücker 2011). Significant statistical heterogeneity will be indicated by a P value on the Chi² test lower than 0.10. Since resilience‐training studies are often conducted with relatively small sample sizes (e.g. Loprinzi 2011; Sood 2014), we acknowledge that the Chi² test has only limited power in such cases. The I² is a descriptive statistic, which equally reflects the percentage of total variation across studies that is due to heterogeneity rather than chance. In accordance with the guidelines of Deeks 2011, we will suppose substantial heterogeneity if an I² is greater than 50%. G² indicates the proportion of unexplained variance, after having allowed for possible small‐study effects (Rücker 2011). No statistical heterogeneity is indicated by a G² near zero.

Assessment of reporting biases

We will assess potential publication bias by inspecting funnel plots (plotting the effect estimates of trials against their standard errors on reversed scales) (Sterne 2011). We acknowledge the fact that funnel plot asymmetry does not necessarily reflect publication bias, but can stem from a number of reasons (Sterne 2011). To differ between real asymmetry and chance, we will follow the recommendations in Sterne 2011 and use Egger’s test (Egger 1997) to test for funnel plot asymmetry, providing there are at least 10 studies included in the meta‐analysis.

Data synthesis

We will synthesise results by describing the resilience interventions, their theoretical concept (when possible), as well as the populations and outcomes studied. We will summarise results in narrative and tabular form. We will perform statistical analyses either in RevMan 2014 or R (R 3.2.2 2015), when appropriate. We will attempt to combine the outcome measures of trials through a pair‐wise meta‐analysis (any resilience training versus control), in order to determine summary (pooled) intervention effects of resilience‐training programmes. The decision to summarise numerical results of RCTs in a pair‐wise meta‐analysis will depend on the number of studies found as well as the heterogeneity of included trials with regard to content or components of resilience interventions, outcomes measured as well as the methodological quality (risk of bias) of selected studies. If intervention studies differ excessively regarding their content, outcomes are too diverse or individual studies are predominantly at high risk of bias, we will not perform a meta‐analysis.

In case a trial reports more than one resilience scale, we will use the scale with better psychometric qualities (as specified in Appendix 3) to calculate effect sizes. If a study reports results for more than one instrument for mental health and well‐being outcomes or for a specific resilience factor, we will select the measure used most often among included studies to calculate effect sizes. In case a study provides data of two instruments used equally frequently in the included RCTs, two review authors (AK, IH) will identify the appropriate measure through discussion (compare Stoffers 2013).

For interventions conducted as preparation for a pre‐defined upcoming stressor or trauma (e.g. military deployment), the stress exposure has to be finished when intervention outcomes are assessed (post‐test or follow‐up) or the stress exposure has to be simulated (e.g. scenarios, video simulation, laboratory stress test) in order to include these studies in the meta‐analysis. This guarantees that the study can be considered as the evaluation of a resilience training and not an intervention fostering related constructs such as mental health.

Since we expect a certain degree of heterogeneity between trials, as indicated by the results of previous reviews (Leppin 2014; Macedo 2014; Robertson 2015), we intend to perform a random‐effects, pair‐wise meta‐analysis using an inverse variance approach, specifically the restricted maximum likelihood method (Veroniki 2015), which is implemented in R (Schwarzer 2015; Viechtbauer 2010). As part of our sensitivity analyses, we will perform both fixed‐effect and random‐effects analyses (see Sensitivity analysis).

Once we have produced a summary of the evidence to date, and only if a pair‐wise meta‐analysis (any resilience training versus control) is possible, we will examine if data are also suitable for a network meta‐analysis (NMA). Network meta‐analyses will be merely exploratory and will only be conducted if the review results in a sufficient and adequate evidence base.

Network meta‐analyses offer the possibility of comparing multiple treatments simultaneously (Caldwell 2005). They combine both direct (head‐to‐head) and indirect evidence (Caldwell 2005; Mills 2012), by using direct comparisons of interventions within RCTs, as well as indirect comparisons across trials on the basis of a common reference group (e.g. an identical control group) (Li 2011). Up to now, a network meta‐analysis on resilience‐training programmes does not exist.

According to Mills 2012, Linde 2016 and the Cochrane Handbook (Higgins 2011e), there are three important conditions for the conduction of NMAs (transitivity, homogeneity, consistency). If a NMA is possible (i.e. the three conditions are fulfilled), we will conduct an analysis ‐ with expert statistical support as suggested by Cochrane (Higgins 2011e) – using a frequentist approach in R (Rücker 2015; Viechtbauer 2015). For sensitivity analyses, the same models will be fitted by the restricted maximum likelihood method (Piepho 2012; Piepho 2014; Rücker 2015). We will consider categorising resilience training into seven groups, based on the underlying training concept: (1) cognitive behavioural therapy, (2) acceptance and commitment therapy, (3) mindfulness‐based therapy, (4) attention and interpretation therapy, (5) problem‐solving therapy, (6) stress inoculation therapy and (7) multimodal resilience training. We may include additional groups after the full literature search has been conducted. Reference groups that will possibly be included in the network meta‐analysis are: attention control, wait‐list, treatment as usual or no intervention. We will investigate inconsistency and flow of evidence in accordance with recommendations in the literature (e.g. Dias 2008; Higgins 2011a; König 2013; Krahn 2013; Krahn 2014; Lu 2006; Lumley 2002; Rücker 2015; Salanti 2008; White 2012a).

Subgroup analysis and investigation of heterogeneity

If substantial heterogeneity is detected, we will examine characteristics of studies that may be associated with this diversity (Deeks 2011). The selection of potential effect modifiers is based on experiences from previous reviews (Leppin 2014; Robertson 2015; Vanhove 2015). We plan to perform the following subgroup analyses:

• setting of resilience interventions (group setting versus individual setting versus combined setting);

• delivery format of resilience interventions (face‐to‐face versus online versus bibliotherapy versus multimodal delivery);

• target group of resilience‐training programmes (employees versus patient populations versus military or police versus general population)³;

• theoretical foundation of resilience‐training programmes (cognitive behavioural therapy versus acceptance and commitment therapy versus mindfulness‐based therapy versus attention and interpretation therapy versus problem‐solving training versus stress inoculation versus multimodal resilience training)³; and

• comparator group in intervention studies (attention control versus wait‐list control versus treatment as usual versus no intervention).

We will only conduct subgroup analyses if we identify 10 or more studies in the review process (Deeks 2011). Moreover, we will restrict the subgroup analyses to our primary outcomes.

³We will provide details in the ‘Differences between protocol and review’ section of the review if the literature search reveals further relevant groups.

Sensitivity analysis

Comparable to the planned subgroup analyses, we will perform sensitivity analyses on the condition that more than 10 RCTs are included in the review. We will also restrict the sensitivity analyses to the primary outcomes.

With regard to intervention studies assessing resilience via resilience scales, we will perform a sensitivity analysis on the basis of the underlying concept (state versus trait) in these measures and limit the analysis to scales assessing resilience as an outcome of an intervention.

In order to examine the impact of the risk of bias of included trials, we will limit the studies to be included in the sensitivity analysis to those whose risk of bias was rated as low or unclear. We will exclude studies assessed at high risk of bias. For studies with low or unclear risk of bias, we will conduct subgroup analyses.

We also plan to consider the restriction to registered studies. We will identify registration both by recording whether we found a study in a trial registry and by noting whether the author claimed to have registered it.

We will perform sensitivity analyses moreover by limiting analysis to those studies with low levels of missing data (less than 10% missing primary outcome). With regards to coping with missing data, we will limit the analysis to studies where missing data were imputed or accounted for by fitting a model for longitudinal data, or where the proportion of missing primary outcome data was less than 10%.

In addition, we intend to check the robustness of our findings, by performing both fixed‐effect and random‐effects analyses in our sensitivity analyses.

We also plan to perform sensitivity analyses based on the ICC estimate in cluster‐randomised trials without adjustment for clustering by excluding cluster‐RCTs where standard errors were not corrected or corrected only on the basis of an externally‐estimated ICC. In an additional sensitivity analysis, we will replace all externally‐estimated ICCs that were less than 0.10, by 0.10.

Finally, we will conduct a sensitivity analysis with regard to the unit of randomisation by limiting the analysis to individually‐randomised trials.