PFI_12mmX178mm.

pdf + eps format

Journal of Child Psychology and Psychiatry 60:10 (2019), pp 1094–1103 doi:10.1111/jcpp.13100

Pathways between early-life adversity and adolescent

self-harm: the mediating role of inflammation in the
Avon Longitudinal Study of Parents and Children
Abigail Emma Russell,1 Jon Heron,1 David Gunnell,1,2 Tamsin Ford,3
4,5
Gibran Hemani, Carol Joinson,1 Paul Moran,1,2 Caroline Relton,4,5
Matthew Suderman,4,5 and Becky Mars1,2
1
Centre for Academic Mental Health, Population Health Sciences, University of Bristol Medical School, Bristol, UK;
2
NIHR Biomedical Research Centre, University Hospitals Bristol NHS Foundation Trust, University of Bristol, Bristol,
UK; 3College of Medicine and Health, University of Exeter, Exeter, UK; 4MRC Integrative Epidemiology Unit,
University of Bristol Medical School, Bristol, UK; 5Population Health Sciences, University of Bristol Medical School,
Bristol, UK

Background: Adverse childhood experiences (ACEs) such as physical and emotional abuse are strongly associated
with self-harm, but mechanisms underlying this relationship are unclear. Inflammation has been linked to both the
experience of ACEs and self-harm or suicide in prior research. This is the first study to examine whether
inflammatory markers mediate the association between exposure to ACEs and self-harm. Methods: Participants
were 4,308 young people from the Avon Longitudinal Study of Parents and Children (ALSPAC), a population-based
birth cohort in the United Kingdom. A structural equation modelling approach was used to fit a mediation model with
the number of ACEs experienced between ages 0 and 9 years old (yo), levels of the inflammatory markers interleukin-
6 and C-reactive protein measured at 9.5 yo, and self-harm reported at 16 yo. Results: The mean number of ACEs
young people experienced was 1.41 (SE 0.03). Higher ACE scores were associated with an increased risk of self-harm
at 16 yo (direct effect relative risk (RR) per additional ACE 1.11, 95% CI 1.05, 1.18, p < 0.001). We did not find
evidence of an indirect effect of ACEs on self-harm via inflammation (RR 1.00, 95% CI 1.00, 1.01, p = 0.38).
Conclusions: Young people who have been exposed to ACEs are a group at high risk of self-harm. The association
between ACEs and self-harm does not appear to be mediated by an inflammatory process in childhood, as indexed by
peripheral levels of circulating inflammatory markers measured in childhood. Further research is needed to identify
alternative psychological and biological mechanisms underlying this relationship. Keywords: Self-harm; suicide;
Avon Longitudinal Study of Parents and Children; adverse childhood experiences; mediation; interleukin-6.
C-reactive protein; Inflammation.

levels of the inflammatory markers C-reactive pro-

Introduction
tein (CRP), interleukin-6 (IL-6) and tumour necrosis
Exposure to adverse childhood experiences (ACEs) is
factor-alpha (TNF-a; Baumeister et al., 2016). This
a well-established risk factor for self-harm (Bj€
orken-
supports data from an earlier systematic review that
stam, Kosidou, & Bj€ orkenstam, 2016; Brown et al.,
found strong evidence for associations between child
2018; Cha et al., 2018; Dube et al., 2001; Hughes
maltreatment and CRP, although the authors noted
et al., 2017; Liu, Scopelliti, Pittman, & Zamora,
that several studies on IL-6 did not find these
2018); however, the psychological and biological
associations (Coelho et al., 2014).
processes underlying this relationship are unclear.
High levels of circulating inflammatory cytokines
Inflammation has been postulated as a potential
indicate systemic inflammation, or chronic activa-
candidate mechanism, as a growing number of
tion of the immune system. Systemic inflammation is
studies have found an association between inflam-
one way through which exposure to ACEs might
matory markers and both ACEs (Baumeister, Akh-
become biologically embedded, whereby psycholog-
tar, Ciufolini, Pariante, & Mondelli, 2016; Coelho,
ical experiences impact on the developmental trajec-
Viola, Walss-Bass, Brietzke, & Grassi-Oliveira,
tory of the immune system with resultant effects on
2014) and self-harm or suicide (Black & Miller,
health (Berens, Jensen, & Nelson, 2017). The bio-
2015). Understanding biological mechanisms
logical inflammatory response in turn triggers a
between ACEs and self-harm is important as it offers
cascade of pathways including affecting levels of
avenues for treatments that could target these
serotonin and dopamine, neurotransmitters known
intermediary pathways in order to reduce the risk
to impact on mood regulation and potentially risk of
of self-harm and suicide.
self-harm (Coelho et al., 2014; Ganguly & Bren-
Findings from a recent systematic review suggest
house, 2015).
that child maltreatment is associated with elevated
Converging lines of evidence from other studies
also support an association between inflammation
Conflict of interest statement: No conflicts declared. and suicidal behaviour (Batty, Bell, Stamatakis, &
© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for Child and
Adolescent Mental Health.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any
medium, provided the original work is properly cited.
doi:10.1111/jcpp.13100 ACEs, inflammation and self-harm 1095

Kivim€ aki, 2016; Black & Miller, 2015). However,

Ethical considerations
most studies are cross-sectional and are not able to
determine the direction of effects between measures Ethical approval for the study was obtained from the ALSPAC
(i.e. whether inflammation precedes suicidal beha- Ethics and Law Committee and the Local Research Ethics
Committees.
viour or vice versa). Evidence for the relationship
between inflammation and self-harm in young peo-
ple is extremely scarce, with a recent review identi- Measures
fying only two studies that have explored this issue
(Kim, Szigethy, Melhem, Saghafi, & Brent, 2014). Primary outcome: self-harm. This was defined as an
affirmative response to the question ‘Have you ever hurt
This is an important omission given that adolescence yourself on purpose in any way (e.g. by taking an overdose of
is a peak time for the onset of self-harm (Hawton, pills, or by cutting yourself)?’ in a questionnaire completed by
Saunders, & O’Connor, 2012). Existing studies have young people at 16 yo.
also predominately been conducted in clinical sam-
ples and have focused on suicide attempts that Exposure: ACEs. Mothers, partners and the study child
comprise only a small proportion of all self-harm were asked 288 questions over 27 data collection points about
episodes (Mars et al., 2014). the child’s exposure to nine ACEs up to 9 yo. ACEs were child
To our knowledge, this is the first study to explore sexual, physical or emotional abuse; parent substance use;
parent mental health problems or suicide attempt; violence
prospective links between ACEs, inflammation and
between parents; parental separation; parental criminal con-
self-harm within the same sample. We used data viction; and child bullying. The first eight are widely used
from a population-based birth cohort to explore (Hughes et al., 2017). We also included bullying, as it has been
whether levels of two key inflammatory markers used as an ACE in other studies (e.g. Finkelhor, Shattuck,
(IL-6 and CRP) mediate the association between Turner, & Hamby, 2013) and is an important negative life event
linked to self-harm (Hawton et al., 2012). Table S1 contains
exposure to ACEs in childhood and self-harm in
definitions of each ACE together with the number of contribut-
adolescence. As early-life adversity is associated ing questions. ACEs were derived as in Houtepen, Heron,
with a range of adverse mental health outcomes Suderman, Tilling, and Howe (2018), and were considered
(Hughes et al., 2017), we also conducted sensitivity present if criteria were met at least once by the time the child
analysis, excluding those with psychiatric disorder, was 9 yo.
to explore whether any effects are independent of
psychopathology. Putative mediators. We explored associations with two
key inflammatory markers (IL-6 and CRP), both of which have
previously been linked to ACEs and to self-harm. IL-6 is a
cytokine that has largely pro-inflammatory effects, and CRP is
an acute phase protein produced by the liver in response to IL-
Methods 6 and TNF-a (Tyrka, Parade, Valentine, Eslinger, & Seifer,
Sample 2015). Blood samples were spun and frozen immediately after
Participants were from the Avon Longitudinal Study of Parents collection at -80°C. IL-6 (pg/L) was measured using high-
and Children (ALSPAC), a population-based birth cohort in the sensitivity IL-6 enzyme-linked immunosorbent assay (R&D
United Kingdom. ALSPAC initially recruited pregnant women Systems, Abingdon, UK). CRP (mg/L) was measured by auto-
resident in Avon, UK, with expected delivery dates between 1 mated particle-enhanced immunoturbidimetric assay (Roche,
April 1991 to 31 December 1992 (Boyd et al., 2013; Fraser UK). Interassay coefficients of variation for both outcomes were
et al., 2012). The initial number of pregnant women who <5%. These methods have been reported in detail in Khandaker
returned at least one questionnaire or attended a ‘Children in et al. (2014). IL-6 and CRP were natural-logarithm-trans-
Focus’ clinic by 19/07/99 was 14,541. Of these initial preg- formed prior to analysis.
nancies, there were a total of 14,676 foetuses, resulting in
14,062 live births and 13,988 children who were alive at 1 year Intermediate confounders and covariates. There
of age. Data were collected via questionnaires and research are four central assumptions relating to confounding in
clinics. The ALSPAC study website contains details of all the mediation analysis. Exposure–outcome confounding, media-
data that are available through a fully searchable data tor–outcome confounding and exposure–mediator confounding
dictionary and variable search tool (http://www.bristol.ac. must be controlled, and no mediator–outcome confounder
uk/alspac/researchers/our-data/). One of each set of twins should be affected by the exposure (VanderWeele, 2016). We
and triplets was included in the current sample. treated child internalising and externalising problems and
The sample for the current study comprised young people body mass index (BMI) as intermediate confounders. These are
who attended the ‘Focus@9’ clinic at approximately 9.5 years factors that may be caused by ACEs and may also causally
old (yo) and provided a blood sample that was successfully contribute to both inflammation and self-harm; that is, they
assayed for IL-6 and CRP. Missing data on exposure, outcome may lie on the causal pathway. SES (maternal education,
and confounders were imputed using chained equations; thus, income and housing tenure), maternal smoking during preg-
our sample was defined by those who had data on the nancy and child sex were treated as confounders of all paths
inflammatory mediators. Individuals reporting illness within (see Appendix S1 for further methodological details).
7 days prior to the blood test were excluded (n = 450) as acute BMI was calculated from young people’s height and weight
illness impacts on levels of inflammatory markers (Ansar & at 9.5 yo. The Strengths and Difficulties Questionnaire (SDQ)
Ghosh, 2013; Khandaker, Pearson, Zammit, Lewis, & Jones, parent-report version, a validated and reliable assessment of
2014), giving a total sample of 4,308. Complete data on dimensional child psychopathology (Goodman, 2001), was
exposure, mediators and outcome variables were available for used to measure internalising and externalising problems
1,619 (the complete case sample; see Figure S1 in the when young people were 8 yo. Both subscales were scored out
Supporting Information for a flow chart detailing sample of 20 with higher scores indicating greater problems (Good-
definition). man, Lamping, & Ploubidis, 2010).

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
1096 Abigail Emma Russell et al. J Child Psychol Psychiatr 2019; 60(10): 1094–103

Smoking during pregnancy was based on maternal self- to calculate relative risk (RR; Cummings, 2009). This can be
report and dichotomised into none or any. Maternal education interpreted as the relative risk of self-harm per additional ACE.
was reported during pregnancy and categorised as no quali- Bootstrapping was used to estimate bias-corrected 95% con-
fications, high school qualifications, and advanced-level or fidence intervals in complete case data (1,000 replications).
college/university qualification. Maternal report of disposable Multiple imputation by chained equations was used to
income was averaged from two questionnaires at 3 and 4 yo. account for missing data (Royston & White, 2011). Our
This was equivalised by scaling according to family size, maximum sample had data on both CRP and IL-6 (N = 4308).
composition and housing benefits and divided into quintiles The study sample was less socioeconomically disadvantaged
(Gregg, Propper, & Washbrook, 2007). Housing tenure during than the wider cohort, with lower levels of psychopathology
pregnancy was dichotomised as rented/housing association (Table S2). ALSPAC has a vast array of auxiliary data that were
versus mortgaged/owned. Child sex was recorded at birth. utilised to make the missing-at-random assumption (MAR)
that underlies multiple imputation plausible (Sterne et al.,
2009). Imputation of each of the study variables used a
Variables for sensitivity analyses bespoke combination of auxiliary data including sociodemo-
graphic variables, earlier and later measures of variables,
We conducted a number of additional sensitivity analyses to measures from a different data collection method/informant,
further explore whether the ACE–self-harm association was and predictors of each variable. Imputation was performed
specific to the age at which self-harm was reported, or whether separately by child sex. Fifty imputed data sets were generated
presence of suicidal intent or psychiatric disorder altered our using the ice command in Stata v15. The main results reported
findings. We examined three secondary outcomes: lifetime are based on the imputed data, with complete case results
history of suicide attempt at 16 yo, multiple episodes of self- reported in the Supporting Information.
harm in past year at 16 yo, and self-harm (regardless of
suicidal intent) at 21 yo. History of suicide attempt at 16 yo
was considered present if young people indicated that they Sensitivity analyses
‘have ever seriously wanted to kill themselves on any occasion
where they have hurt themselves’ or the ‘last time [they] hurt We conducted sensitivity analyses with secondary dichoto-
themselves it was because [they] wanted to die’. As our primary mous outcomes of suicide attempt at 16 yo, multiple self-harm
outcome would also capture those who had self-harmed only at 16 yo and self-harm at 21 yo. We also restricted our model
once, we used multiple self-harm in a sensitivity analysis to to those without psychiatric disorder at 15 (7.5% excluded).
represent those with more chronic self-harm. Young people Finally, we excluded those with CRP values of >10 mg/L, as
were asked about the frequency of self-harm within the past some have argued that values over this threshold are indicative
year; we dichotomised this into none or once versus more than of acute infection/immune activation (Ansar & Ghosh, 2013).
once. Self-harm at 21 yo was measured the same way as the A similar threshold or cut-off has not been proposed or used for
primary outcome. Sensitivity analyses explored whether effects IL-6. We examined whether findings were due to imprecision in
were independent of psychopathology by excluding those with measures of inflammatory markers by repeating our mediation
psychiatric disorder, measured using the Development and analysis with a latent inflammation variable as the mediator,
Well-Being Assessment (DAWBA) at 15 yo. DSM-IV diagnosis carried out in MPlus version 8 (Muth en & Muthen, Los Angeles,
was generated using information from parents and young CA) using Monte Carlo integration and unimputed data
people via a computer algorithm (Goodman, Ford, Richards, (n = 1,811).
Gatward, & Meltzer, 2000; Goodman, Heiervang, Collishaw, &
Goodman, 2011).
An additional marker of systemic inflammation, the DNA
methylation neutrophil-to-lymphocyte ratio (mdNLR), was also Results
available for a subsample of 1,000 ALSPAC children (Relton Descriptive results (imputed data)
et al., 2015). DNA methylation data and cell type algorithms
were used to calculate the ratio (Koestler et al., 2017). During Approximately one quarter of the sample reported
systemic inflammation, neutrophil counts rise while lympho- self-harm at 16 yo (24.5%, 95% CI 22.17, 26.83).
cyte counts drop; higher mdNLR values therefore indicate The mean number of ACEs young people experi-
greater inflammation. mdNLR was derived from DNA methyla-
enced was 1.41 (95% CI 1.35, 1.47). The distribution
tion profiles as in Ambatipudi et al. (2018). It was natural-
logarithm-transformed prior to analysis and utilised in a latent of ACE scores was positively skewed, with a median
variable comprised of mdNLR, IL-6 and CRP. of 1 (Figure 2). The frequency of each ACE and their

Analysis
We utilised the ACE score as a continuous variable as a
likelihood ratio test showed that a categorical model was not a
better fit (v2(2) = 2.50, p = 0.29). Tetrachoric correlations
between ACEs were reported, and Poisson regression of each
ACE on the outcome, controlling for covariates, was con-
ducted. A mediation model was fitted to determine whether the
association between ACEs from birth to 9 yo and self-harm at
16 yo was mediated by inflammatory markers at 9.5 yo. As IL- Figure 1 Directed acyclic graph (DAG) of mediation model. Notes:
6 is upstream of CRP in the inflammatory response, both were ‘C’ covariates: maternal smoking during pregnancy, child sex,
included in one model (Figure 1). We used a generalised maternal education, income, housing tenure. ACEs, adverse
structural equation modelling approach to partition the asso- childhood experiences; SDQ, Strengths and Difficulties Question-
ciation between ACE score, IL-6, CRP and self-harm into naire; BMI, body mass index. Each intermediate confounding
indirect and direct effects with robust standard errors using path (BMI, SDQs) was specified separately, shown as one for
the gsem and nlcom commands in Stata v15 (StataCorp LLC, clarity in figure. IL-6, interleukin-6; CRP, C-reactive protein. ACEs
College Station, TX) (Gunzler, Chen, Wu, & Zhang, 2013). As were measured from 0 to 9 years of age, IL-6 and CRP at
self-harm was not a rare outcome, we used Poisson regression 9.5 years, and self-harm at 16 years

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.13100 ACEs, inflammation and self-harm 1097

association with self-harm is shown in Table 1. The Table 1 Exposure to adverse childhood experiences (ACEs)
most frequently reported ACE was parent mental and associations between each ACE and self-harm at 16 yo
(imputed data N = 4,308)
health problems or suicide attempt (39.3%). Violence
between parents (21.7%) and parental separation Adverse
(21.6%) were the next most common. Sexual abuse childhood Per
was the least experienced, with a frequency of less experience cent 95% CI RR 95% CI p
than 1% (complete case results are shown in Sexual abuse 0.8 0.48, 1.19 1.21 0.53, 2.77 0.65
Table S3). Of the individual ACEs, the strongest Physical abuse 7.6 6.67, 8.58 1.26 0.96, 1.65 0.09
evidence for an association with self-harm was found Emotional 19.1 17.6, 20.5 1.32 1.07, 1.62 0.01
for emotional abuse (RR = 1.32, 95% CI 1.07, 1.62) abuse
and parental separation (RR = 1.27, 95% CI 1.06, Parent 11.7 10.5, 12.9 1.15 0.88, 1.50 0.30
substance use
1.54; Table 1). ACEs were highly correlated with one Parent mental 39.3 37.7, 41.0 1.17 0.99, 1.38 0.07
another, except for sexual abuse and bullying health
(Table S4). Table 2 shows descriptive statistics for problems or
the imputed and complete case samples. suicide
In regression models exploring associations attempt
Violence 21.7 20.1, 23.4 1.18 0.93, 1.50 0.16
between the exposure, mediators and outcome, between
ACE score was associated with both self-harm and parents
with IL-6. There was tentative evidence of an asso- Parental 21.6 20.1, 23.0 1.27 1.06, 1.54 0.01
ciation between IL-6 and self-harm; however, the separation
confidence interval included the null value. There Child 12.7 11.6, 13.8 1.25 0.99, 1.58 0.06
experiences
was no evidence of an association between ACEs and bullying
CRP, or CRP and self-harm (Table S5). Parent criminal 6.6 5.68, 7.43 1.05 0.74, 1.48 0.80
conviction

Mediation results For definitions of each adverse childhood experience, see

Table S1. Fifty imputed data sets. CI, confidence interval;
In the mediation model, higher ACE scores were RR, relative risk. Associations between each ACE and self-
associated with an increased risk of self-harm at harm modelled in Poisson regression controlling for child sex,
16 yo, such that with each additional ACE, an maternal smoking during pregnancy, income, maternal edu-
individual was 11% more likely to report self-harm cation and housing tenure.
(direct effect RR 1.11, 95% CI 1.05, 1.17, p < 0.001).
We did not find evidence of an indirect effect via IL-6
and CRP, indicating that this association was not psychiatric disorder and when excluding those indi-
mediated by inflammation at 9.5 yo (indirect effect viduals with CRP > 10 mg/L (Table 3). When explor-
RR 1.00, 95% CI 1.00, 1.01, p = 0.38; Table 3, ing associations with suicide attempts at 16 yo, the
Figure S2). These results were consistent in sensi- RR was higher than in the main analysis that
tivity analyses with multiple self-harm at 16 yo, self- included both suicidal and nonsuicidal self-harm
harm assessed at 21 yo, when excluding those with (total effect RR 1.22, 95% CI 1.11, 1.33, p < 0.001),
but this association was not mediated through IL-6
and CRP. Finally, we treated inflammation as a
latent variable, incorporating the mdNLR, and found
no evidence of mediation (Table S6). Findings were
consistent across the imputed and complete case
samples.

Discussion
We investigated inflammation as a potential mecha-
nism linking exposure to early-life adversity and
adolescent self-harm. An association was found
between the total number of ACEs and self-harm,
with each additional ACE conferring an eleven per
cent increase in risk. The magnitude of these effects
is in line with other studies (Dube, Felitti, Dong,
Giles, & Anda, 2003) and confirms the established
association between adversity and self-harm found
Figure 2 Number of adverse childhood experiences per child in the literature (Liu et al., 2018). We extend prior
(imputed data N = 4,308). Notes: 50 imputations. Due to small ns,
work in this area by exploring whether this relation-
those experiencing five or more adverse childhood experiences
are grouped [Colour figure can be viewed at wileyonlinelibrary.c ship was mediated by two key inflammatory mark-
om] ers, IL-6 and CRP; however, we did not find evidence

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
1098 Abigail Emma Russell et al. J Child Psychol Psychiatr 2019; 60(10): 1094–103

Table 2 Descriptive statistics of study variables, imputed data There are several plausible explanations for our
(N = 4,308) and complete case sample (N = 1,619) findings. First, it is possible that the timing of our
Imputed
measures may have had an impact on our ability
data Complete case to detect associations. Much of the literature on
(N = 4,308) sample (N = 1,619) ACEs and inflammation focuses on adulthood,
Variable Mean SE n Mean SD
with few studies conducted in childhood (Baldwin
et al., 2018; Danese et al., 2011; Slopen, Kubzan-
ACE score 1.41 1.28 1,619 1.11 1.24 sky, McLaughlin, & Koenen, 2013). The immune
IL-6 (pg/ml) 1.20 1.45 1.16 1.32 system continues to mature throughout adoles-
CRP (mg/L) 0.62 1.95 0.57 1.48
BMI age 9 17.6 0.04 17.4 2.55
cence, and it may be that differences in systemic
SDQ internalising 2.84 0.04 1,596 2.68 2.58 inflammation due to ACEs may not manifest until
problems adulthood (Kuhlman, Chiang, Horn, & Bower,
SDQ externalising 4.77 0.06 1,596 4.28 3.19 2017). Our ACE measure was comprehensive,
problems derived from 288 questions, asked between birth
and 9 yo; however, there could be sensitive peri-
Total
Variable % SE n n % ods during which exposure to ACEs has a greater
impact on the developing immune system. Evi-
Self-harm age 16 24.5 1.19 1,619 304 18.8 dence for sensitive periods comes from a study
Suicide attempt age 16 12.0 1.23 1,619 98 6.1
that used slightly different measures of adversity
Multiple self-harm age 16 16.2 1.31 1,578 159 10.8
Self-harm age 21 23.0 1.14 1,174 245 20.9 and explored their association with CRP in
Child sex (female) 48.9 0.76 1,619 891 55.0 ALSPAC. ACEs experienced from 6 to 8 yo were
Housing tenure (not 17.1 0.58 1,594 150 9.4 associated with CRP; ACEs experienced from 1.5
owned/mortgaged) to 6 yo were not, however, unless included in a
Maternal education
cumulative index from 0 to 8 yo (Slopen et al.,
Degree 16.7 0.57 1,608 411 25.6
A-level 26.9 0.68 500 31.1 2013). Alternatively, associations may differ
GCSE 35.0 0.74 504 31.3 according to the type of ACE experienced, with
<GCSE 21.5 0.64 193 12.0 more biologically salient ACEs such as physical
Equivalised household 1,562 abuse having stronger effects (Kuhlman et al.,
income (quintiles)
2017). Investigating whether results are impacted
Highest (1) 22.1 0.67 466 29.8
2 21.3 0.66 392 25.1 by the timing or type of ACE was beyond the
3 21.5 0.66 323 20.7 scope of the current study, but is an important
4 19.3 0.64 241 15.4 area for future research.
Lowest (5) 15.9 0.60 140 9.0 Research interest in the ACE–inflammation asso-
Maternal smoking during 19.8 0.61 1,619 211 13.0
ciation is increasing. A recent study on victimisation
pregnancy (ever)
Psychiatric disorder age 15 7.5 0.56 1,433 59 4.1 in childhood and adolescence found associations
with inflammatory markers by age 18 in females but
SE, standard error; CI, confidence interval; SD, standard not males (Baldwin et al., 2018). The construct of
deviation; ACE, adverse childhood experiences; IL-6, inter- victimisation overlaps with several of our ACEs
leukin-6; CRP, C-reactive protein; BMI, body mass index; SDQ,
Strengths and Difficulties Questionnaire (internalising and
including bullying, exposure to domestic violence,
externalising scores out of 20 with higher scores indicating physical and sexual abuse. The relationship between
more problems). Psychiatric disorder age 15 by Development victimisation and inflammation may depend on the
and Well-Being Assessment (DAWBA) computer algorithm. type and timing of victimisation. For example, asso-
Fifty imputed data sets were generated. ciations between adolescent but not childhood sex-
ual abuse and adult CRP levels have been reported
while the same was not found for physical abuse
for mediation via inflammation in this sample. We (Bertone-Johnson, Whitcomb, Missmer, Karlson, &
found evidence of a univariate association between Rich-Edwards, 2012). Peer victimisation in child-
ACEs and IL-6, but not CRP. This lack of association hood and adolescence has been associated with
with CRP is in line with findings from a recent meta- higher levels of CRP in midlife (Takizawa, Danese,
analysis that found no significant association Maughan, & Arseneault, 2015) and also in young
between adversity and CRP, with evidence being adulthood (Copeland et al., 2014). In addition, peer
weakest for studies in middle childhood (Kuhlman, victimisation within the past week has been found to
Horn, Chiang, & Bower, 2019). A growing number of be associated with an increased inflammatory
studies have demonstrated the existence of a link response to immune challenges in adolescent girls
between ACEs and inflammation, and between (Giletta et al., 2018). These findings demonstrate
inflammation and self-harm behaviours (Baumeister that the temporal relationship between ACEs and
et al., 2016; Coelho et al., 2014; Kim et al., 2014; inflammation is complex, with ACEs having both
Serafini et al., 2013), highlighting this as a suitable acute and chronic effects on the immune system.
candidate mechanism. However, ours is the first Our measure of ACEs only captured experiences by
study to provide a direct test of this hypothesis. the age of nine and our measure of inflammatory

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.13100 ACEs, inflammation and self-harm 1099

Table 3 Results of mediation model; association between adverse childhood experiences and self-harm via inflammatory markers
interleukin-6 and C-reactive protein (imputed data N = 4,308)

Indirect effect via IL-6

Direct effect and CRP Total effect

Model RR 95% CI p RR 95% CI p RR 95% CI p

Main analysis
Self-harm at age 16 1.11 1.05, 1.17 <0.001 1.00 1.00, 1.00 0.380 1.11 1.05, 1.18 <0.001
Sensitivity analyses
Self-harm at age 21 1.12 1.06, 1.18 <0.001 1.00 1.00, 1.01 0.617 1.12 1.06, 1.18 <0.001
Multiple self-harm at age 16 1.10 1.03, 1.17 0.007 1.00 1.00, 1.00 0.668 1.10 1.03, 1.17 0.007
Suicide attempt at age 16 1.22 1.11, 1.33 <0.001 1.00 1.00, 1.01 0.806 1.22 1.11, 1.33 <0.001
Excluding those with psychiatric disorder at 1.11 1.05, 1.17 <0.001 1.00 1.00, 1.01 0.313 1.11 1.05, 1.17 <0.001
age 15 (n = 4,138)
Excluding those with CRP values of >10 mg/ 1.11 1.05, 1.17 <0.001 1.00 1.00, 1.00 0.318 1.11 1.05, 1.18 <0.001
L excluded (n = 4,278)

All models adjusted for BMI, internalising and externalising problems as intermediate confounders, and child sex, maternal
smoking during pregnancy, income, maternal education and housing tenure as covariates; CRP, C-reactive protein. Fifty imputed
data sets generated. RR, relative risk; CI, confidence interval.

markers was taken at a much younger age the inflammation–suicidal behaviour association
(9.5 years) than in these studies (largely in adult- comes from large studies utilising national registries
hood), which may explain why we did not find the of healthcare records (Serafini et al., 2013). Cases of
same association. self-harm in these studies will only be recognised
Another possibility is that the mechanism through when an individual seeks treatment. Those who
which inflammation impacts on self-harm does not present to services differ from those who do not, in
occur through systemic inflammation, as assessed terms of demographic characteristics, as well as
in this study, but rather as an altered inflammatory severity and method of self-harm (Hawton et al.,
response to immune system challenges. Indeed, we 2012).
did not find evidence that IL-6 or CRP levels pre- A further possible explanation for our findings
dicted self-harm. Some previous studies have not is that that there is no causal relationship
found an association between circulating levels of between ACEs and self-harm operating through
inflammatory markers and ACEs but have instead inflammation. Indeed, a causal association
reported differential IL-6 responses to both psycho- between ACEs and self-harm is yet to be empir-
logical and bacterial challenges (Miller & Chen, ically established. Two recent studies have
2010). demonstrated that part of the association between
A third possibility is that prior studies may be adversity and self-harm is attributable to genetic
detecting inflammatory consequences of self-harm confounding, but also that exposure to adversity
as opposed to antecedents, as the act of self-harm confers an additional risk of self-harm, indicating
itself will trigger an immune response and thus a small causal influence of adversity on self-harm
inflammation. Much of the evidence for the inflam- (Baldwin et al., 2019; Richmond-Rakerd et al.,
mation–self-harm relationship comes from cross- 2019). A myriad of social, psychological and
sectional studies in adult populations, which are biological factors have been proposed to play a
unable to establish the direction of association role in the aetiology of self-harm (see Hawton
(Brundin, Erhardt, Bryleva, Achtyes, & Postolache, et al., 2012). To rule out the causal role of
2015). Our study is also novel in that we explored inflammation, genetically informed methods such
associations with self-harm regardless of suicidal as Mendelian randomisation should be used to
intent, whereas prior research has focused on supplement existing knowledge. If there is no
suicidal behaviour. This is important as suicide support for a causal association in either direc-
attempts make up only a small proportion of total tion, further research should focus on other
self-harm episodes (Mars et al., 2014). Sensitivity potential causal pathways to elicit potential tar-
analysis focusing only on suicide attempts found a gets for screening, prevention or intervention.
similar pattern of results to our main findings, yet
the total effect was stronger for the suicide attempt
Strengths and limitations
outcome than for the main analysis, with the
relative risk of self-harm increasing to 22% per Our study is the first to explore the association
additional ACE. between ACEs, inflammation and self-harm. Data
A further difference between our study and previ- were from a large population-based birth cohort with
ous research is the use of a population-based information on over 4,000 young people. This is
sample. The strongest epidemiological evidence for important, as most cases of self-harm are not known

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
1100 Abigail Emma Russell et al. J Child Psychol Psychiatr 2019; 60(10): 1094–103

to services (Hawton et al., 2012). Our ACEs were who have been affected may be beneficial for
prospectively assessed across childhood, and the preventing self-harm behaviours. Future studies
longitudinal design enabled us to clearly establish are needed to explore the impact of ACE type, and
the direction of effects between our variables. We timing of exposure, on self-harm risk. Further
also adjusted for a range of confounders and were research should also explore alternate biological
also able to explore associations with several differ- and psychological pathways through which ACEs
ent measures of inflammation, including IL-6, CRP might influence risk of self-harm and suicide.
and the mdNLR. Plausible candidates include HPA axis dysregula-
Our findings need to be interpreted in the light of tion, puberty and epigenetic modifications (Patton
several limitations. First, nonresponse and loss to et al., 2007; Turecki & Brent, 2016).
follow-up have been shown to occur more fre-
quently among individuals with particular charac-
teristics (including socioeconomic disadvantage, Supporting information
ACEs and psychopathology; Wolke et al., 2009) Additional supporting information may be found online
that may lead to bias. In the current study, young in the Supporting Information section at the end of the
people who did not consent to give blood had article:
higher SDQ emotional problems scores than those
who provided consent (data available on request). Appendix S1. Confounders in mediation analysis.
Our study sample was less socioeconomically dis- Figure S1. Participant flow chart.
advantaged and experienced less psychopathology Figure S2. Path diagram of mediation results.
than the wider ALSPAC cohort, which may limit the Table S1. Adverse childhood experiences (ACEs).
generalisability of our findings. Under the missing- Table S2. Descriptive statistics, study sample and
at-random assumption, multiple imputation will ALSPAC core sample.
correct for any biases present in the complete case Table S3. Adverse childhood experience frequencies in
analyses, and our findings using imputed data complete case study sample (N = 1619).
were comparable with the complete case. We also Table S4. Tetrachoric correlations between adverse
childhood experiences (unimputed data, n = 2446).
conducted a number of sensitivity analyses, and
Table S5. Univariable and adjusted associations
our findings were unchanged, increasing confi-
between mediation variables (imputed data N = 4308).
dence in our conclusions.
Table S6. Complete case sensitivity analyses mediation
Second, ACEs were reported by mothers and
results.
fathers/partners (with the exception of bullying)
and may have led to an underestimate in preva-
lence, particularly for ACEs related to abuse. We Acknowledgements
also did not include neglect as an ACE. Neglect This study was funded by the Medical Research Founda-
was measured by ALSPAC; however, no suitable tion and the Medical Research Council (Grant ref: MR/
questions had been asked by 9 yo, the cut-off we R004889/1) Pathways to self-harm: Biological mecha-
used to ensure that our measure of ACEs was prior nisms and genetic contribution (PI: BM). M.S., G.H. and
to the measure of inflammatory markers, and thus, C.R. work in the Medical Research Council Integrative
we decided not to include it in the current study. Epidemiology Unit at the University of Bristol which is
Self-harm was self-reported by young people at 16 supported by the Medical Research Council and the
and 21 yo and may be subject to misreporting University of Bristol (Grant Numbers MC_UU_00011/1,
(Mars et al., 2016). However, estimates of the MC_UU_00011/4 and MC_UU_00011/5).
The UK Medical Research Council and Wellcome
prevalence in our complete case data (18.8%) are
(Grant ref: 102215/2/13/2) and the University of
in line with international estimates (Gillies et al.,
Bristol provide core support for ALSPAC. This publica-
2018). tion is the work of the authors, and A.R. and B.M. serve
In addition, it is possible that ACEs occurring after as guarantors for the contents of this paper. A compre-
the age of 9 would increase the risk of self-harm at hensive list of grant funding is available on the ALSPAC
16 yo. However, because of the nature of our medi- website (http://www.bristol.ac.uk/alspac/external/d
ation model, these effects would not have been ocuments/grant-acknowledgements.pdf); this research
captured by the pathway that we measured. This was specifically funded by the Wellcome Trust (Grant
would weaken our ability to detect an indirect effect ref: GR067797MA).
via inflammation at 9.5 yo. Further studies could D.G., P.M., C.R., and B.M. are supported by the NIHR
control for later ACE exposure or explicitly model Biomedical Research Centre at University Hospitals
Bristol NHS Foundation Trust and the University of
this within the analysis.
Bristol, England. The views expressed in this publica-
In conclusion, we found ACEs between the ages
tion are those of the author(s) and not necessarily those
of 0 and 9 were associated with an increased risk of the NHS, the National Institute for Health Research
of adolescent self-harm, but the association was or the Department of Health and Social Care.
not mediated by markers of inflammation mea- The authors are extremely grateful to all the families
sured in late childhood. Strategies aimed at pre- who took part in this study, the midwives for their help
venting ACEs and interventions targeted at those in recruiting them, and the whole ALSPAC team, which

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.13100 ACEs, inflammation and self-harm 1101

includes interviewers, computer and laboratory techni-

cians, clerical workers, research scientists, volunteers, Correspondence
managers, receptionists and nurses. The authors would Abigail Emma Russell, Centre for Academic Mental
like to thank Lotte Houtepen for her work on the ACE Health, Population Health Sciences, Bristol Medical
constructs in ALSPAC. The authors have declared that School, Oakfield House, Oakfield Grove, Bristol, BS8
they have no competing or potential conflicts of interest. 2BN, UK; Email: a.e.russell@bristol.ac.uk

Key points

 Adverse childhood experiences (ACEs) are a strong predictor of self-harm; however, the mechanisms
underlying this association are unclear. Inflammation has been linked to both ACEs and self-harm in separate
studies.
 In the first study of its kind, we explored whether systemic inflammation, indexed by interleukin-6 and C-
reactive protein levels, mediated the association between ACEs and adolescent self-harm.
 We found a strong association between total number of ACEs and self-harm, with each additional ACE
conferring an additional 11% risk of self-harm at age 16.
 We did not find evidence that this association was mediated by inflammation in childhood, suggesting that
inflammatory markers may not be a useful biomarker for self-harm risk among those exposed to early-life
adversity.

Black, C., & Miller, B.J. (2015). Meta-analysis of cytokines and

References chemokines in suicidality: Distinguishing suicidal versus
Ambatipudi, S., Sharp, G.C., Clarke, S.L.N., Plant, D., Tobias, nonsuicidal patients. Biological Psychiatry, 78, 28–37.
J.H., Evans, D.M., . . . & Relton, C.L. (2018). Assessing the Boyd, A., Golding, J., Macleod, J., Lawlor, D.A., Fraser, A.,
role of DNA methylation-derived neutrophil-to-lymphocyte Henderson, J., . . . & Davey, Smith G. (2013). Cohort profile:
ratio in rheumatoid arthritis. Journal of Immunology The ‘children of the 90s’—the index offspring of the Avon
Research, 2018, 2624981. Longitudinal Study of Parents and Children. International
Ansar, W., & Ghosh, S. (2013). C-reactive protein and the Journal of Epidemiology, 42, 111–127.
biology of disease. Immunologic Research, 56, 131–142. Brown, R.C., Heines, S., Witt, A., Braehler, E., Fegert, J.M.,
Baldwin, J.R., Arseneault, L., Caspi, A., Fisher, H.L., Moffitt, Harsch, D., & Plener, P.L. (2018). The impact of child
T.E., Odgers, C.L., . . . & Danese, A. (2018). Childhood maltreatment on non-suicidal self-injury: Data from a rep-
victimization and inflammation in young adulthood: A resentative sample of the general population. BMC Psychi-
genetically sensitive cohort study. Brain, Behavior, and atry, 18, 181.
Immunity, 67, 211–217. Brundin, L., Erhardt, S., Bryleva, E., Achtyes, E.D., & Posto-
Baldwin, J.R., Arseneault, L., Caspi, A., Moffitt, T.E., Fisher, lache, T. (2015). The role of inflammation in suicidal
H.L., Odgers, C.L., . . . & Danese, A. (2019). Adolescent behaviour. Acta Psychiatrica Scandinavica, 132, 192–203.
victimization and self-injurious thoughts and behaviors: A Cha, C.B., Franz, P.J., M Guzm an, E., Glenn, C.R., Kleiman,
genetically sensitive cohort study. Journal of the E.M., & Nock, M.K. (2018). Annual Research Review: Suicide
American Academy of Child & Adolescent Psychiatry, 58, among youth–epidemiology,(potential) etiology, and treat-
506–513. ment. Journal of Child Psychology and Psychiatry, 59, 460–
Batty, G.D., Bell, S., Stamatakis, E., & Kivim€ aki, M. (2016). 482.
Association of systemic inflammation with risk of completed Coelho, R., Viola, T., Walss-Bass, C., Brietzke, E., & Grassi-
suicide in the general population. JAMA Psychiatry, 73, Oliveira, R. (2014). Childhood maltreatment and inflamma-
993–995. tory markers: A systematic review. Acta Psychiatrica Scan-
Baumeister, D., Akhtar, R., Ciufolini, S., Pariante, C.M., & dinavica, 129, 180–192.
Mondelli, V. (2016). Childhood trauma and adulthood Copeland, W.E., Wolke, D., Lereya, S.T., Shanahan, L.,
inflammation: A meta-analysis of peripheral C-reactive pro- Worthman, C., & Costello, E.J. (2014). Childhood bullying
tein, interleukin-6 and tumour necrosis factor-a. Molecular involvement predicts low-grade systemic inflammation into
Psychiatry, 21, 642. adulthood. Proceedings of the National Academy of Sciences
Berens, A.E., Jensen, S.K., & Nelson, C.A. (2017). Biolog- of the United States of America, 111, 7570–7575.
ical embedding of childhood adversity: From physiolog- Cummings, P. (2009). The relative merits of risk ratios and
ical mechanisms to clinical implications. BMC Medicine, odds ratios. Archives of Pediatrics & Adolescent Medicine,
15, 135. 163, 438–445.
Bertone-Johnson, E.R., Whitcomb, B.W., Missmer, S.A., Karl- Danese, A., Caspi, A., Williams, B., Ambler, A., Sugden, K.,
son, E.W., & Rich-Edwards, J.W. (2012). Inflammation and Mika, J., . . . & Arseneault, L. (2011). Biological embedding of
early-life abuse in women. American Journal of Preventive stress through inflammation processes in childhood. Molec-
Medicine, 43, 611–620. ular Psychiatry, 16, 244–246.
Bj€
orkenstam, E., Kosidou, K., & Bj€ orkenstam, C. (2016). Dube, S.R., Anda, R.F., Felitti, V.J., Chapman, D.P., Wil-
Childhood household dysfunction and risk of self-harm: liamson, D.F., & Giles, W.H. (2001). Childhood abuse,
A cohort study of 107 518 young adults in Stockholm household dysfunction, and the risk of attempted suicide
County. International Journal of Epidemiology, 45, throughout the life span: Findings from the Adverse Child-
501–511. hood Experiences Study. JAMA, 286, 3089–3096.

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
1102 Abigail Emma Russell et al. J Child Psychol Psychiatr 2019; 60(10): 1094–103

Dube, S.R., Felitti, V.J., Dong, M., Giles, W.H., & Anda, R.F. Kim, J.-W., Szigethy, E.M., Melhem, N.M., Saghafi, E.M., &
(2003). The impact of adverse childhood experiences on Brent, D.A. (2014). Inflammatory markers and the patho-
health problems: Evidence from four birth cohorts dating genesis of pediatric depression and suicide: A systematic
back to 1900. Preventive Medicine, 37, 268–277. review of the literature. The Journal of Clinical Psychiatry,
Finkelhor, D., Shattuck, A., Turner, H., & Hamby, S. (2013). 75, 1242–1253.
Improving the adverse childhood experiences study scale. Koestler, D.C., Usset, J., Christensen, B.C., Marsit, C.J.,
JAMA Pediatrics, 167, 70–75. Karagas, M.R., Kelsey, K.T., . . . & Wiencke, J.K. (2017).
Fraser, A., Macdonald-Wallis, C., Tilling, K., Boyd, A., Golding, DNA methylation-derived neutrophil-to-lymphocyte ratio:
J., Davey Smith, G., . . . & Lawlor, D.A. (2012). Cohort profile: An epigenetic tool to explore cancer inflammation and
The Avon Longitudinal Study of Parents and Children: outcomes. Cancer Epidemiology, Biomarkers & Prevention,
ALSPAC mothers cohort. International Journal of Epidemiol- 26, 328–338.
ogy, 42, 97–110. Kuhlman, K.R., Chiang, J.J., Horn, S., & Bower, J.E. (2017).
Ganguly, P., & Brenhouse, H.C. (2015). Broken or maladap- Developmental psychoneuroendocrine and psychoneuroim-
tive? Altered trajectories in neuroinflammation and behavior mune pathways from childhood adversity to disease. Neuro-
after early life adversity. Developmental Cognitive Neuro- science & Biobehavioral Reviews, 80, 166–184.
science, 11, 18–30. Kuhlman, K.R., Horn, S.R., Chiang, J.J., & Bower, J.E. (2019).
Giletta, M., Slavich, G.M., Rudolph, K.D., Hastings, P.D., Early life adversity exposure and circulating markers of
Nock, M.K., & Prinstein, M.J. (2018). Peer victimization inflammation in children and adolescents: A systematic
predicts heightened inflammatory reactivity to social stress review and meta-analysis. Brain, Behavior, and Immunity.
in cognitively vulnerable adolescents. Journal of Child Psy- https://doi.org/10.1016/j.bbi.2019.04.028
chology and Psychiatry, 59, 129–139. Liu, R.T., Scopelliti, K.M., Pittman, S.K., & Zamora, A.S.
Gillies, D., Christou, M.A., Dixon, A.C., Featherston, O.J., (2018). Childhood maltreatment and non-suicidal self-in-
Rapti, I., Garcia-Anguita, A., . . . & Christou, P.A. (2018). jury: A systematic review and meta-analysis. The Lancet
Prevalence and characteristics of self-harm in adolescents: Psychiatry, 5, 51–64.
Meta-analyses of community-based studies 1990–2015. Mars, B., Cornish, R., Heron, J., Boyd, A., Crane, C., Hawton,
Journal of the American Academy of Child & Adolescent K., . . . & Gunnell, D. (2016). Using data linkage to investigate
Psychiatry, 57, 733–741. inconsistent reporting of self-harm and questionnaire non-
Goodman, R. (2001). Psychometric properties of the strengths response. Archives of Suicide Research, 20, 113–141.
and difficulties questionnaire. Journal of the American Mars, B., Heron, J., Crane, C., Hawton, K., Kidger, J., Lewis,
Academy of Child & Adolescent Psychiatry, 40, G., . . . & Gunnell, D. (2014). Differences in risk factors for
1337–1345. self-harm with and without suicidal intent: Findings from
Goodman, R., Ford, T., Richards, H., Gatward, R., & Meltzer, the ALSPAC cohort. Journal of Affective Disorders, 168,
H. (2000). The development and well-being assessment: 407–414.
Description and initial validation of an integrated assess- Miller, G.E., & Chen, E. (2010). Harsh family climate in early
ment of child and adolescent psychopathology. The Journal life presages the emergence of a proinflammatory phenotype
of Child Psychology and Psychiatry, 41, 645–655. in adolescence. Psychological Science, 21, 848–856.
Goodman, A., Heiervang, E., Collishaw, S., & Goodman, R. Patton, G.C., Hemphill, S.A., Beyers, J.M., Bond, L., Toum-
(2011). The ‘DAWBA bands’ as an ordered-categorical mea- bourou, J.W., McMORRIS, B.J., & Catalano, R.F. (2007).
sure of child mental health: Description and validation in Pubertal stage and deliberate self-harm in adolescents.
British and Norwegian samples. Social Psychiatry and Journal of the American Academy of Child & Adolescent
Psychiatric Epidemiology, 46, 521–532. Psychiatry, 46, 508–514.
Goodman, A., Lamping, D.L., & Ploubidis, G.B. (2010). When Relton, C.L., Gaunt, T., McArdle, W., Ho, K., Duggirala, A.,
to use broader internalising and externalising subscales Shihab, H., . . . & Smith, G. (2015). Data resource profile:
instead of the hypothesised five subscales on the Strengths Accessible resource for integrated epigenomic studies (ARIES).
and Difficulties Questionnaire (SDQ): Data from British International Journal of Epidemiology, 44, 1181–1190.
parents, teachers and children. Journal of Abnormal Child Richmond-Rakerd, L.S., Trull, T.J., Gizer, I.R., McLaughlin, K.,
Psychology, 38, 1179–1191. Scheiderer, E.M., Nelson, E.C., . . . & Martin, N.G. (2019).
Gregg, P., Propper, C., & Washbrook, E. (2007). Understanding Common genetic contributions to high-risk trauma expo-
the relationship between parental income and multiple child sure and self-injurious thoughts and behaviors. Psycholog-
outcomes: A decomposition analysis. LSE STICERD ical Medicine, 49, 421–430.
Research Paper No. CASE129. Royston, P., & White, I.R. (2011). Multiple imputation by
Gunzler, D., Chen, T., Wu, P., & Zhang, H. (2013). Introduction chained equations (MICE): Implementation in Stata. Journal
to mediation analysis with structural equation modeling. of Statistical Software, 45, 1–20.
Shanghai Archives of Psychiatry, 25, 390–394. Serafini, G., Pompili, M., Seretti, M.E., Stefani, H., Palermo,
Hawton, K., Saunders, K.E., & O’Connor, R.C. (2012). Self-harm M., Coryell, W., & Girardi, P. (2013). The role of inflamma-
and suicide in adolescents. The Lancet, 379, 2373–2382. tory cytokines in suicidal behavior: A systematic review.
Houtepen, L.C., Heron, J., Suderman, M.J., Tilling, K., & European Neuropsychopharmacology, 23, 1672–1686.
Howe, L.D. (2018). Adverse childhood experiences in Slopen, N., Kubzansky, L.D., McLaughlin, K.A., & Koenen,
the children of the Avon Longitudinal Study of Parents K.C. (2013). Childhood adversity and inflammatory pro-
and Children (ALSPAC). Wellcome Open Research, 3, cesses in youth: A prospective study. Psychoneuroen-
106. docrinology, 38, 188–200.
Hughes, K., Bellis, M.A., Hardcastle, K.A., Sethi, D., Butchart, Sterne, J.A., White, I.R., Carlin, J.B., Spratt, M., Royston, P.,
A., Mikton, C., . . . & Dunne, M.P. (2017). The effect of Kenward, M.G., . . . & Carpenter, J.R. (2009). Multiple
multiple adverse childhood experiences on health: A sys- imputation for missing data in epidemiological and clinical
tematic review and meta-analysis. The Lancet. Public Health, research: Potential and pitfalls. BMJ, 338, b2393.
2, e356–e366. Takizawa, R., Danese, A., Maughan, B., & Arseneault, L.
Khandaker, G.M., Pearson, R.M., Zammit, S., Lewis, G., & (2015). Bullying victimization in childhood predicts inflam-
Jones, P.B. (2014). Association of serum interleukin 6 and mation and obesity at mid-life: A five-decade birth cohort
C-reactive protein in childhood with depression and psy- study. Psychological Medicine, 45, 2705–2715.
chosis in young adult life: A population-based longitudinal Turecki, G., & Brent, D.A. (2016). Suicide and suicidal
study. JAMA Psychiatry, 71, 1121–1128. behaviour. The Lancet, 387, 1227–1239.

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.13100 ACEs, inflammation and self-harm 1103

Tyrka, A.R., Parade, S.H., Valentine, T.R., Eslinger, N.M., & Wolke, D., Waylen, A., Samara, M., Steer, C., Goodman, R.,
Seifer, R. (2015). Adversity in preschool-aged children: Ford, T., & Lamberts, K. (2009). Selective drop-out in
Effects on salivary interleukin-1b. Development and Psy- longitudinal studies and non-biased prediction of behaviour
chopathology, 27, 567–576. disorders. The British Journal of Psychiatry, 195, 249–256.
VanderWeele, T.J. (2016). Mediation analysis: A practitioner’s
guide. Annual Review of Public Health, 37, 17–32. Accepted for publication: 25 June 2019

© 2019 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.