Background

The term “Adverse Childhood Experiences” or “ACEs” was first coined in Felitti et al.’s [1] seminal “Adverse Childhood Experiences Study” and was used to describe a group of specific childhood experiences. Adverse childhood experiences (ACEs) can broadly be defined as potentially traumatic life events occurring in the first 18 years of life [2]. Experiences that are defined as ACEs vary within the literature; however, they can broadly be categorised into three overarching classifications: abuse (emotional, sexual, and physical), neglect (emotional and physical), and household dysfunction (alcohol and/or drug abuse in the house, imprisoned family member, mother treated violently, and parental loss, separation, or divorce) [3]. While these ACEs are the most heavily researched, this list is not exhaustive. There are further experiences recognised as ACEs in research that this review will also consider, such as being bullied [4], community and collective violence [5], parental mortality and morbidity [6], child marriage, [7] and child trafficking [8].

Over the last three decades, extensive research has explored the relationship between ACEs and the later onset of poorer cognitive, emotional, and behavioural outcomes [1, 9]. Strong cross-sectional and longitudinal relationships have been established between ACEs and an increased risk of developing various psychiatric problems including depression [10], anxiety disorders [11], suicidal ideation [12] and psychosis [13]. The increasing body of extant literature has concluded that ACEs are a dangerous public health problem [14], and emerging research has recognised adult mental illness as one of the largest public financial burdens associated with ACEs [15].

Adverse childhood experiences and poor life outcomes

In 1998, Felitti et al. conducted the aforementioned “Adverse Childhood Experiences Study” in Southern California in the United States of America. The retrospective cohort study collected data over two waves from 1995–1997 and was responded to by 17,337 participants. Participants were selected for the study from their attendance of the Kaiser Permanente’s Health Appraisal Centre (HAC) due to being adult members of the Kaiser Health Plan in San Diego County. The study was designed to explore whether there was a relationship between early life adversity and adult physical and mental ill-health. In both waves, adults who had completed a standard medical evaluation at HAC one-to-two weeks’ prior were asked about adverse childhood experiences (ACEs) and health behaviours through questionnaires sent by mail. The HAC evaluations provided standardised medical histories and formed part of the ACE study database. There were ten adverse childhood experiences included that were separated into two broader categories of childhood maltreatment and household dysfunction [16]: emotional and physical neglect; emotional, sexual, or physical abuse; living in a household where members abused substances, where there was violence against the mother, where members were mentally ill/ suicidal or where members were ever incarcerated; and parental separation or divorce.

Felitti et al. [1] found that around two-thirds of the sample experienced at least 1 ACE and around 12.5% experienced at least 4 ACEs. When exploring later negative life events, the researchers found a variety of health outcomes that were strongly associated with having 4 or more ACEs. For example, compared to having no ACEs, those with 4 or more were around 4.6 times more likely to have had depressed mood in the past year, 12.2 times more likely to have ever attempted suicide, 7.4 times more likely to consider themselves and alcoholic and 10 times more likely to have ever injected drugs. A strong dose–response relationship was established between one’s number of ACEs and poor outcomes, including the emergence of later life mental difficulties and physical diseases.

After the pioneering work of Felitti et al. [1], ACEs studies have been conducted globally that confirm ACEs are associated with a variety of poor outcomes [17]. For example, studies have evidenced the association between ACEs and suicidal behaviour in South Africa [18], heavy drinking amongst other health-harming behaviours in the United Kingdom [19], depressed affect in California, North America [20], illicit drug use in Brazil, South America [21] and anxiety, depression and PTSD symptoms in South-East Asia [22]. In recent years, ACE studies have also been synthesised in systematic reviews and meta-analyses. In their systematic review and meta-analysis, Hughes et al. [23] demonstrated the significant, deleterious effect multiple ACEs have on lifelong health. Other systematic reviews include Norman et al. [24] and Kalmakis and Chandler [25], whose results suggested significant associations between ACEs and various long-term mental health outcomes and health-harming behaviours, including depressive disorders, suicide attempts, PTSD, substance misuse, and sexual risk behaviour. Sahle and colleagues’ [26] recent umbrella review also confirmed strong, significant associations between ACEs and common mental disorders.

Rationale

Despite the seminal ACE study [1] following the original participants to measure the emergence of poor health outcomes over time, the study still measured ACEs retrospectively. In current literature, retrospective reporting of ACEs by adults remains the most common method of obtaining comprehensive self-reports of adversity [27]. Studies using test–retest reliability to explore the consistency of reports of ACEs over time generally find stability in retrospective measures [28]. However, due to the reporting of adversity being many years after the event occurred [29], one must consider the possible biases that may result in inaccurate data. Scepticism of the validity of childhood information collected in adulthood has existed for over five decades now, as Yarrow, Campbell and Burton [30] suggested recollection of childhood information may be largely contingent on the information and narration of events told by one’s parents. Retrospective reporting of ACEs is thought to be at a far higher risk of inaccuracy than prospective reporting (the reporting of ACEs as they emerge) due to further issues such as recall bias [31], memory decay [32] and mood-congruent bias [33], where the reporting of historical events is determined by one’s current mental state. For example, researchers have posited that adults diagnosed with mental disorders such as depression exhibit specific “retrieval biases” that subsequently result in superior recall of more negative historical events and fewer positive events [34, 35].

Henry, Moffitt, Caspi, Langley and Silva [36] explored the agreement between retrospective and prospective reporting of ACEs across a prospectively studied large sample of adolescents. Several categories of information were compared and whilst more objective content such as moving house and height were consistently reported between prospective and retrospective measures, the poorest agreement was found in the more subjective information such as one’s psychological state and childhood adversities such as maternal mental illness and family conflict. The lack of agreement between retrospective and prospective reports of childhood adversities has also been substantiated in more recent research. For example, Baldwin et al.’s [37] systematic review and meta-analysis found that around 52% of participants who prospectively reported adversity in childhood did not go forward to report it retrospectively. Furthermore, 56% of participants who retrospectively disclosed ACEs had not reported this adversity prospectively. Whilst it has been argued the poor agreement between retrospective and prospective approaches to reporting is due to poor validity of the retrospective measures, there may be other reasons for such disagreement. For example, prospective measures may record ACEs before childhood ends and subsequently may not capture adverse events that happened after data collection in the way that retrospective accounts of adversity across the whole of childhood do [37]. This current systematic review has subsequently chosen to only include studies using prospective measures of ACEs in line with Baldwin et al.’s [37] recommendation not to compare studies across prospective and retrospective approaches to data collection. This is primarily due to the large discrepancy in populations they identify.

The current review will include prospective, longitudinal research designs that study ACEs instead of retrospective, cross-sectional designs due to their ability to explore temporal sequencing of events [38]. Prospective studies offer valuable information about developmental changes, incidence rates of ACEs, and a better understanding of the timing and chronicity of ACEs [39, 40]. Furthermore, without the temporal patterning of events, the direction of the relationships cannot not be established [41]. This is one of the main reasons why retrospective adult studies of ACEs are not sufficient to understand causal pathways between ACEs and adult outcomes [42]. In prospective longitudinal studies, the collection of data through time allows opportunity for confounding variables to be measured and adjusted for at each time point [43]. However, it should be acknowledged that causal mechanisms between adverse childhood experiences and later-life poor outcomes such as mental ill-health are difficult to infer- even in longitudinal research [44]. This is due to many factors including under-reporting biases in the reporting of ACEs [39] and a lack of consideration of unobservable genetic components and family characteristics that contribute to any causal relationships [44]. These limitations mean we do not aim to infer any causal relationships from the findings in our review. Despite the limitations of prospective longitudinal ACEs studies, prospective measures of ACEs still provide a valuable tool for identifying risk markers for later poor outcomes in adults [45]. The six mental health outcomes (depression, anxiety, PTSD, suicidal ideation, self-harm, and psychotic-like experiences) were selected as they represent six of the most commonly assessed mental health outcomes in research exploring the association between ACEs and later-life mental ill-health.

Methods/ design

Aim and review questions

The main aim of this systematic review and meta-analysis is to address the gap in the literature by exploring the associations between ACEs and the specific adult mental ill-health outcomes of depression, anxiety, PTSD, psychotic-like experiences, suicidality, and self-harm in prospective longitudinal research globally. A considerable portion of prospective longitudinal research focuses on the relationship between ACEs and mental health outcomes earlier in development (e.g., [46,47,48,49,50]). However, we are interested in exploring whether such associations between ACEs and mental health remain into adulthood and across the lifespan. There have been less syntheses of such longer-term associations, and this was a main reason we wanted to limit our review to adult mental health outcomes.

The authors are aware of a similar systematic review and meta-analysis that recently explored longitudinal associations between childhood trauma and adult mental disorder [51]. However, the current review provides the novel inclusion of grey literature, differing mental health outcomes (unlike McKay et al. [51] who included the outcomes of depression, anxiety, psychotic disorder and bipolar disorder, this study seeks to include anxiety, depression, psychotic-like experiences, PTSD, suicidality, and self-harm) and a lower threshold for the measurement of mental health outcomes. Unlike McKay et al. [51], the current study stipulates the mental health outcomes need not be formal psychiatric diagnoses using established diagnostic criteria for mental disorders in adulthood as such use of these measures is rare in low-and middle-income countries. Furthermore, this review completes an updated and more comprehensive database search (including ProQuest Dissertations and Theses comprising of grey literature), which, in turn, may reduce potential effects of algorithm or publication bias [52]. We felt that to ensure we captured a holistic overview of all literature on the topic that grey literature should be included. Grey literature is often excluded from large systematic reviews, and we feel that this may unintentionally exclude certain geographical locations that lack funding to support peer-reviewed study production and publication. The Newcastle–Ottawa Scale will still be used to appraise study quality of any grey literature found and their findings would still have to fit our stringent inclusion and exclusion criteria.

This protocol has been registered in PROSPERO (CRD42021297882) and followed the PRISMA-P (Preferred Reporting Items for Systematic Review andMeta-Analysis Protocols) 2015 statement: recommended items to address in a systematic review protocol [53] (see checklist in Additional file 1).

Certain questions may not be answered as they remain contingent on enough studies fitting the criteria. The Population-Issue-Comparison-Outcome (PICO)/ Population-Exposure-Outcome (PEO) framework [54] was used to create the overarching review question which is:

“Are adults who have been exposed to adversity in childhood at an increased risk of developing mental illness(es) compared to adults who have not been exposed to adversity in childhood?”

We will address the following sub-questions:

  1. 1.

    What are the associations between ACEs and depression, anxiety, PTSD, suicidal ideation, self-harm, and psychotic-like experiences in adulthood with a specific interest in the prevalence of research conducted in high-income countries versus low-and middle-income countries?

  2. 2.

    Which geographical locations does the evidence on ACEs stem from?

  3. 3.

    Which ACEs have the largest negative associations with adult mental health?

  4. 4.

    Is there a cumulative effect of ACEs on mental health outcomes?

  5. 5.

    Is the association between ACEs and adult mental ill-health moderated by geographical location of study?

  6. 6.

    Is the association between ACEs and adult mental ill-health moderated by peer-reviewed status?

  7. 7.

    Is the association between ACEs and adult mental ill-health moderated by study design or analysis?

  8. 8.

    Is the association between ACEs and adult mental ill-health dependent on age of onset at the first adversity?

  9. 9.

    What is the quality of studies looking at longitudinal associations between ACEs and mental health outcomes?

Question 4 pertains to any study that includes a measure of cumulative adversity. We aim to pool the effect sizes from analyses that have used, for example, a continuous measure of cumulative ACEs such as “0 Adversity, 1 Adversity, 2 Adversity” or a measure using the widely recognised cut-off 0–3 ACEs vs 4 + ACEs. Question 5 was created as the geographical location of studies may influence the prevalence and types of childhood adversity. For example, evidence suggests ACEs may be more common in low-resource settings/ low- and middle-income countries [55,56,57]. Question 6 was included as there may be publication bias present. This means the studies published in peer-reviewed journals could over-represent the significance of associations given the fact that many articles published show statistically significant results [58, 59]. Question 7 was created as analytic choices across studies may influence the results found and reported [60]. Furthermore, the effect sizes reported for the relationships between ACEs and adult mental health may vary depending on what design is used (e.g., whether the study used a self-reported, prospective measure of adversity or whether the study used data-linkage to official court records). This may be due to many reasons, including under-reporting biases and different thresholds of what events are recorded or “counted” as an adversity [61,62,63]. Question 8 was created as the age of onset may align with critical/ sensitive periods for the development of mental health symptoms and thus could again influence the strength of associations [64].

Inclusion and exclusion criteria

We adopted the Population-Exposure-Outcome (PEO) model to aid in outlining the inclusion and exclusion criteria seen in Tables 1 and 2, respectively.

Table 1 Inclusion criteria using the PEO model where applicable
Full size table
Table 2 Exclusion criteria using the PEO model where applicable
Full size table

Information sources

For this review, twelve electronic databases will be searched: Embase, PsycINFO, MEDLINE (Ovid version), and Global Health through the Ovid interface. ProQuest will be used to search Public Affairs Information Service (PAIS), Dissertations and Theses, Sociology Database (including Sociological Abstracts and Social Services Abstracts), PTSDpubs (formerly PILOTS) and ASSIA. CINAHL, WHO Global Index Medicus, and WHO Violence Info will also be searched. The search was conducted throughout the month of June, 2021. An update of the search from June 2021 until March 2023 will also be carried out before the full review to ensure the review includes the most up-to-date research. The search will be limited to publication dates from 1990 onwards and to human subjects in databases that include this limiter. This specific period has been chosen as it aligns with the drafting of the United Nations Convention on the Rights of the Child (UNCRC) by the United Nations [65]. It should be noted that studies published after 1990 that used data from cohorts prior to 1990 will still be eligible if all inclusion criteria are satisfied. This has been decided as the study rationale, research design, research questions, analyses and findings will be interpreted with knowledge from the UNCRC, including a universal definition of when childhood ends and detailed conceptualisations of child protection and maltreatment [66]. The English language specification will be manually screened.

To ensure literature saturation, the authors of this review will email authors of known large cohort studies in the relevant field of research to query whether they have any research that is unfinished/ in the process of being published. Search terms can be found in Appendix 1 and a table of definitions of key concepts can be found in Appendix 2.

Search strategy

Examples of the search strategies can be found in the Appendices 3, 4, 5, 6, 7 and 8. The search strategy will be altered to account for varying syntax, limiters, and expanders in different databases.

Data management

Studies identified by the database searches will be extracted and be uploaded to Covidence (a systematic review management software). Before importing search results into Covidence, database citations and abstracts will be exported into Zotero where they will be de-duplicated. Then, references will be transformed into a RIS file format. Once imported to Covidence, duplicates will be checked for and removed again.

Selection and collection process: screening and extraction

Abstracts and titles will be independently double screened to determine whether the studies meet the inclusion criteria. Next, the remaining papers will be subject to a full-text screen for assessment of inclusion by two reviewers. If necessary, additional information will be sought from the authors of included studies. Any discrepancies in the decision to include a study in the final review will be resolved by team discussion or a third independent reviewer. The final review will include a PRISMA flow diagram documenting the flow of studies throughout the systematic review process.

The final data extracted from the remaining studies will be stored in a spreadsheet on Covidence. The data extracted by reviewers will include:

  • General study information (First author, year of study, the format that the information is presented in (e.g., book, article, thesis, conference proceeding)).

  • More specific study characteristics (Study location, sample size, sample source (e.g., cohort name), study design (e.g., birth cohort or data linkage), numbers exposed to ACE and outcome).

  • Sociodemographic information of participants (gender, age, socio-economic status, ethnicity).

  • Information about study variables (measurement/ tool(s) used to collect ACE and mental health data, type of ACEs measured, source of ACEs reporting, type of mental health outcomes measured, age adversity/ mental health was recorded at).

  • Information regarding the analysis (metrics, adjustments, results).

Risk of bias (quality) assessment

Study quality (evaluated in review question 9) will be assessed using the Newcastle–Ottawa Scale for cohort studies and case–control studies (NOS) [67]. This assessment of quality implements a star system based on three overarching domains of study characteristics: Selection of Study Groups, Comparability of Groups and Ascertainment of Exposure/ Outcome. Typically, a maximum of 8 stars can be awarded (A maximum award of 1 star per item within the domains Selection and Exposure and a maximum award of 2 stars for the domain of Comparability) [68]. Two reviewers will independently assess the methodological quality of the included studies and any discrepancies in agreement will be resolved by a third reviewer. However, we will not give each included study an overall quality score or “total star rating”. This is in line with limitations of overall quality scores highlighted in the Cochrane Handbook for Systematic Review of Interventions [69], including a lack of uniformity of quality appraisals across different quality scales being largely attributable to differing conceptualisations of “quality”.

Data synthesis

A narrative synthesis of included studies will be completed with study information presented in tables and in text. The qualitative discussion will include tabular summaries of the included studies and a discussion of the relationships within and between the studies and will answer review questions 1–4. If enough studies are identified by the database searches and they have enough similarity in design, multi-level meta-analyses will be conducted using the “metafor” [70] package in R to answer review questions 1 and 3–8. The meta-analysis will implement a random-effects model as it is predicted reported effect sizes will vary as a function of exposure, the measurement tools used, and differences in the populations from which the samples are drawn. Specifically, odds ratios (ORs) will be computed in the meta-analysis and when the study outcome is a continuous measure, Hasselblad and Hedges’ [71] method will be used to convert standardised mean differences to log odds ratios. ORs have been cited as a preferred computation for effect size over risk ratio (RR) when computing meta-analyses with binary data (see [72,73,74]). This is given odds ratios’ symmetry regarding outcome definition and their homogenous, constant nature [75]. The minimum number of studies to permit meta-analyses is two studies per mental health outcome. Again, if enough studies permit, meta-regressions will be conducted in which the moderating effects of the age of adversity onset, country, types of adversity, publication status, and duration of follow-up period will be explored.

I2 will be used to assess statistical heterogeneity. It was originally intended to be independent of the number of studies (unlike Cochran’s Q) and has been regularly used in Cochrane reviews [76]. However, it should be noted some research suggests I2 can still be biased in small meta-analyses [77].

Meta bias(es)

The possibility of publication/ dissemination bias in the identified studies will be explored. Publication bias will be identified and corrected by first using the “trim and fill” method [78] which will be conducted for each outcome in the meta-analysis. This procedure will help detect and correct any asymmetry in the funnel plots. The Egger bias test will be computed for further examination of funnel plot asymmetry [79].

Discussion

The purpose of this review is to systematically investigate the existence and strength of association between ACEs and adult mental health outcomes in prospective longitudinal studies with a focus on the mental health outcomes anxiety, depression, PTSD, self-harm, suicidal ideation, and psychotic-like experiences.

First, by exploring associations between ACEs and key mental health outcomes, we aim to evaluate the importance of identifying prospectively measured individual ACEs and cumulative ACE scores as risk markers for later poor mental health outcomes in adults [45]. Second, by exploring how ACEs relate to different mental health outcomes, we may assist in the future prioritisation of specific preventative mental health interventions in ACE-exposed populations. Third, we will also evaluate whether studies in the field of childhood adversity are affected by publication bias. This will provide further insight as to whether the included published studies are a representative sample of available evidence of the longitudinal associations between ACEs and adult mental health. Lastly, this review may have further implications for ACEs research such as identifying methodological weaknesses and knowledge gaps in literature that can be addressed in future primary studies. For example, we may be able to tell what ACEs and mental health outcomes are under-researched and whether there are regions of the world that are under-represented or missing from the literature.

The authors acknowledge the risk of bias that results from being unable to include studies not readily available in English. Whilst this decision was made due to resource constraints, authors may miss high-quality studies and key data [80]. We must also consider limitations associated with the use of official records (e.g., child protective service records or court cases) to obtain information about ACE exposure in prospective ACE studies. Official records are more likely to include only the most severe cases of childhood adversity and are more likely to document ACEs that happened chronically or earlier in life [81]. They subsequently miss childhood experiences that may not require official child protective services record such as childhood bullying or parental divorce, but that may still be significantly associated with poor outcomes [82, 83]. Furthermore, prospectively measured ACEs may also be vulnerable to under-reporting due to substantiation bias, report bias, investigation bias, and issues relating to stigma and secrecy [84,85,86]. Despite the limitations outlined, prospective measures of ACEs provide valuable information about temporal patterning of ACEs and later-life mental ill-health.

In conclusion, studies exploring longitudinal associations between ACEs and adult mental health outcomes have already been synthesised, but this review aims to expand the existing systematic review methodological and analytical approaches. We aim to offer valuable insights about the associations between ACEs and mental health outcomes, their moderators, the quality of longitudinal ACEs studies, specific methodological weaknesses and knowledge gaps that may influence future research directions such as targeting under-researched locations, ACEs, and mental health outcomes.