Population and exposure(s).

Included studies will have a population composed of at least one primary caregiver (G1) who experienced a traumatic event and their child (G2) and/or grandchild (G3) who did not experience this traumatic event firsthand. The G1 individual would not need to directly report their traumatic exposure, as long as there is some method to assess this, such as indirect reporting by G2. The traumatic event in G1 must have occurred before the birth of G2. G1 or G2 parents do not need to be biologically related to the child of interest. We will include trauma-exposed parents regardless of their trauma-related symptomatology. Animal studies will be excluded.

A traumatic event will be defined as exposure to threatened death, serious injury, or sexual violence, as outlined in the Diagnostic and Statistical Manual of Mental Disorders [24]. We will also consider parental trauma-related symptoms as the exposure, contingent that these symptoms are a direct result of an eligible traumatic exposure.

Mediator(s) and moderator(s).

Included studies will have at least one psychosocial factor operating at the family-level (including household- and parent-levels) that measured the interpersonal relationship between the child and their immediate family and/or primary caregiver before the child has reached 18 years of age. These could be framed or examined as mediators or moderators. Original articles would not need to explicitly use these terms or conceptualize factors in this way to be eligible.

Psychosocial factors might include parental psychiatric symptoms, suggesting that parental trauma-related mental health symptoms could be considered as the exposure or the mechanism, depending on the context. Factors might also include parental psychological factors, interpersonal parent-child relationships, or social factors within the household-level (which include demographic factors) that would influence child-rearing behaviors [25]. Comparators will be defined the measurement of the absence of the factor(s) or varying levels of continuously-measured factor(s) of interest in at least some sub-group of participants in the study sample. We will also include demographic factors (for example, age, gender, socioeconomic status, etc.) as moderators.

Outcome(s).

Included studies will quantify either the presence or absence of direct or indirect effects of trauma in the child (G2 or G3) under observation. Given the heterogeneity in clinical presentations of intergenerationally-transmitted trauma, we will only include quantitatively measured markers of psychological symptoms/functioning (i.e., child psychological development, psychopathology, and/or indicators of wellbeing). To increase our yield, we will be inclusive in what outcomes would qualify as child psychopathology and/or indicators of wellbeing for screening and may restrict this criterion after we assess the final sample of included studies. Outcomes will be considered in offspring. Psychiatric symptoms in the child would not need to be directly observed; for example, it would be acceptable for G1 to report on G2 symptoms.

Date and language restrictions.

This review will have no date restrictions, but will be limited to studies published in English, French, Kinyarwanda, and Spanish languages, for feasibility, as research team members are fluent in these languages.

Study design and document type.

We will include any epidemiologic study design, such as cross-sectional studies, case-control studies, cohort studies, randomized controlled trials, or the quantitative component of mixed-methods studies. Systematic review papers, meta-analyses, case-studies, qualitative studies, and study protocols will be excluded.

Due to feasibility, we will only include published literature (peer-reviewed literature and dissertations). One exception is that we will include conference abstracts during the screening phase contingent upon gaining access to the full text document (see Source of evidence selection). We included conference abstracts and dissertations to reduce publication bias.

Mediation meta-analyses.

We will use a mediation meta-analysis approach to pool data from individual mediation studies that report enough statistical information to derive indirect, direct, and/or total effects at the path-level. This could include reporting of beta coefficients directly (standardized or not standardized), or through back-calculating these estimates using correlation matrices, means, and variances/standard deviations for the exposure or traumatic exposure/symptoms (x), mediator of interest (m), and outcome or mental health status (y) variables. Our primary analyses will assess the following pathway: parental trauma exposure → psychosocial mechanisms (including parental trauma-related symptoms) → child mental health outcome. We will also assess the following pathway: parental trauma-related symptoms → psychosocial mechanisms → child mental health outcome. We will conduct several meta-analyses according to each mediating construct and whether it mediated G1 to G2, G2 to G3, or G1 to G3 effects. We will conduct sensitivity analyses to test whether we could consolidate findings of individual studies with different study designs and consolidate individual studies that measured the exposure as parental trauma versus parental post-traumatic symptoms. We plan to merge outcomes together to assess the global effect on child psychopathology; if there is enough data, we will sub-divide outcomes into different meta-analytic models. We will conduct meta-analyses following guidelines reported in the literature [29–31] that would be most suitable to for the methods reported in our sample of included studies, depending on the distributions and availability of statistical information reported in the included studies. We will aim to pool data as long as there are at least two beta estimates available for an eligible path.

Moderation meta-analyses.

We will aim to use a moderation meta-analysis approach to pool moderator data from individual studies, using similar methods as described above. Our primary analyses will assess moderating effects on parental trauma exposure → child mental health outcome; our secondary analyses will assess moderating effects on parental trauma-related symptoms → child mental health outcome. We will perform moderation meta-analyses using standard statistical methods [32], though we will take an atypical analytic approach given the nature of our research question. Traditional moderation meta-analyses assess the impact of study-level factors on the degree of association between an exposure and outcome [33, 34] and are restricted to assessing moderators that apply to every participant within analyses. For example, this could include moderating factors such as study design (such as whether cross-sectional studies versus cohort studies moderate the exposure-outcome relationship) or other population-level characteristics (such as study location or public versus private funding source). For the purpose of this review, however, it is unlikely this meta-analytic approach would be feasible to understand the moderating role of psychosocial family-level factors on the intergenerational transmission of trauma, as it is unlikely that individual studies would restrict their analyses to different levels for these types of factors, and thus, it would be unlikely that we would retrieve enough studies to assess differences in effect sizes using this traditional approach. For example, if we aimed to understand the moderating effect of parental communication style, it is unlikely that individual studies assessing the relationship between parental trauma exposure and child mental health outcome would have restricted their analyses to one specific level of communication style. To increase our likelihood of pooling estimates, and provide more meaningful narrative syntheses, we instead plan to combine individual studies that directly assessed for moderation as part of their analyses, as reported elsewhere [35]. We would aim to pool this data using an integrative data analysis approach (otherwise known as individual participant data meta-analysis) and would assess moderation using joint tests of interaction using linear regression models [33], though this will be contingent on securing original datasets from the included studies. Should it be challenging to obtain the individual-level datasets, we will report moderation findings using narrative syntheses.

Assessment of methodological quality.

We will assess the quality of mediation paths using the quality assessment tool used in other mediation meta-analyses [15]. To our knowledge, this is the only tool available to evaluate the quality of mediation studies. To assess the quality of moderation paths, we will assess the quality of included articles using the JBI critical appraisal tools, selected based on the study design of the included paper [36, 37]. We will apply this framework at the study level. We selected this tool as, to our knowledge, there are no tool developed specifically to evaluate the quality of moderation analyses. If we gather enough studies, we will conduct sensitivity analyses to assess whether study methodological quality impacts pooled estimates.