An alarming proportion of young people in the UK report being bullied. In one study, we found that 30% of over 110,000 15-year olds in the UK report regular bullying (two to three times a month or more) in the last couple of months (Przybylski and Bowes 2017). That alone is a problem: being bullied is horrible, and no child should be subjected to it.

Numerous studies, ranging from retrospective studies of clinical populations to longitudinal analyses of large birth cohorts, have found that childhood bullying is associated with depression, anxiety, self-harm, suicidal ideation and attempts and other poor mental health and social outcomes (Copeland, Wolke et al. 2013, Wolke, Copeland et al. 2013, Wertz 2016). So we have a substantial proportion of young people exposed to a stressor that’s associated with a wide range of poor mental health outcomes; as argued in a recent Mental Elf blog, we need to do more to tackle bullying (Ford, 2018).

Who gets bullied?

Understanding why some young people may be more likely to be bullied than others might help us develop better prevention programs. Individual characteristics seem to matter: low self-esteem, depressive symptoms and ADHD have all been associated with an increased risk of being bullied (Arseneault, Bowes et al. 2010). Factors beyond the individual are also important, including child maltreatment to socioeconomic disadvantage (Bowes, Arseneault et al. 2009).

At the moment, all of this research evidence comes from classical observational studies. A key weakness is that we can never be certain that we are adjusting for all the possible factors that might account for the relationship; unmeasured confounding always remains. We know from other studies that children’s genes seem to matter; twin studies indicate the being bullied is heritable (Ball, Arseneault et al. 2008). But these kinds of twin studies can’t tell us which genes are associated with an increased risk.

Then there is the chicken and egg problem. Is it that bullying leads to depressive symptoms, or was it some early manifestation of those same depressive symptoms that led to the child being bullied in the first place (Brugger 2017)? As we can never be completely confident that we have measured the timing of variables exactly, reverse causation is usually a problem. Enter the novel research design used by Schoeler and colleagues: the multi-polygenic risk score approach. Say what?!

Methods

Participants

Schoeler and colleagues used data from 5,028 participants of the ALSPAC study.

Outcome

Children self-reported their experiences of being bullied at three different time points when they were aged 8, 10 and 13 years. The study authors took the mean average across all three years as their study outcome. So much, so simple.

Predictors

Here’s where it gets interesting: the authors derived 35 polygenic risk scores (PGSs), based on discovery genome-wide-association studies (GWAS) for these different traits. The PGSs were computed for a range of mental health difficulties (including depression and anxiety), traits related to cognition (intelligence), personality (e.g. neuroticism) and physical measures (e.g. Body Mass Index). They also looked at ‘negative controls’ to check that PGSs for traits that shouldn’t be associated with risk of bullying (like osteoporosis) weren’t.

Why does it matter whether a ‘polygenic risk score’ predicts bullying?

A polygenic risk score aggregates the effects of many common genetic variants associated with any given trait into a single individual-level score. If depression, for example, is causally associated with a child’s likelihood of being bullied, then the polygenic risk score for depression (a genetic proxy of depressive symptoms) should predict whether or not the child was bullied. And this time, unlike in previous studies, reverse causation is unlikely to pose much of a problem.

Analysis

The study authors first tested whether the 35 PGSs individually predicted bullying, and then looked at the multivariate associations to look at the unique effect of each PGS.

Results

Individual associations between each PGS and being bullied

In the first step, 11 different PGS scores were associated with bullying:

• The strongest correlations were present for cognitive measures like intelligence and educational attainment, and measures relating to mood, like depressive symptoms and neuroticism.
• Interestingly, PGSs for loneliness and anxiety weren’t associated, and neither was the PGS for Autism.

Independent associations in multivariate analyses

In the final multivariate stage, 5 PGSs were independently associated with exposure to bullying:

1. Diagnosis of depression
2. ADHD
3. Risk-taking
4. BMI
5. Intelligence

Combining all the PGSs together, children in the lowest quintile had a predicted prevalence of bullying of 15.7%, compared to 36.6% among children in the highest quintile.

There was no evidence for any differences by sex, and the results remained similar when the authors controlled for the chronicity of bullying experienced.

Conclusions

Genetic proxies predict who gets bullied

This study indicates that having a genetic predisposition to certain individual vulnerabilities and traits is associated with risk of being bullied. Some of these vulnerabilities already have a robust evidence base, particularly depressive symptoms and ADHD. Some risk factors were more surprising, with genetic predispositions to risk-taking and having a higher BMI being associated with an increased risk of being bullied; a genetic proxy for having a higher intelligence was associated with a reduced risk of being bullied. No other genetic proxy was related to bullying, including genetic proxies for other mental health disorders (e.g. bipolar disorder and obsessive compulsive disorder), personality traits (e.g neuroticism), autism, and physical measures other than BMI (e.g. height).

Strengths and limitations

The strengths of this research lie in the cool research design: by using genetic proxies for individual risk factors for bullying, this study design helps to overcome some of the problems with previous observational studies, namely unmeasured confounding and reverse causation. But no research design is perfect; as noted by the study authors, there are some important assumptions, the key ones being that the genetic proxy is associated with the outcome only through its effects on the exposure (e.g. the genetic score for ADHD is associated with exposure to bullying only through its effects on ADHD symptoms) and that the genetic proxy is independent of all factors that confound the association between exposure and outcome. In a multi-polygenic model such as that used by the study authors, the risks of pleiotropy (i.e. that the genetic proxy affects more than one trait) goes up.

There are also some other important limitations about the data itself. For me, perhaps the biggest one is that the measure of bullying is self-report. There is always a challenge here: kids who are more prone to depression may be more likely to remember and report negative events (like bullying) compared to kids without these symptoms. They’re also more likely to display ‘hostile attributional biases’; interpreting neutral social actions as being more hostile and nasty than other kids. So we can’t rule out the possibility here that the genetic proxy for depression is associated with a greater perception of reporting of bullying, and not necessarily of more actual, objective bullying.

The sample is also non-representative; for purposes of the genetic analysis, all individuals with non-European ancestry (including self-reported) were removed. This relates to a wider issue in the field; the underrepresentation of many diverse populations in biomedical sciences.

Drop out from longitudinal studies is always an issue, and in this case it means that the most genetically vulnerable children may not have been included in analysis (though the authors point out the children who dropped out had similar family histories of psychiatric disorders and early developmental characteristics as those retained in the study).

Finally, the authors also focus on individual traits, which clearly are important, but we shouldn’t overlook the fact that social factors like exposure to maltreatment and social disadvantage also affect who gets bullied (Bowes, Arseneault et al. 2009). Genetic risks operate in a particular context. For example, in this study, a polygenic score for high BMI was correlated with risk of being bullied. In cultures where high BMI does not lead to ridicule and shame, this would likely not be the case.

Implications for practice

This study provides further evidence that certain individual characteristics increase children’s risk of being bullied. Whilst whole-school anti-bullying programs have been shown to be effective at reducing bullying (Ttofi and Farrington 2011), these results also suggest possible avenues for prevention programs. Educating children about early mental health difficulties and reducing stigma may help to reduce the risk of children with depressive symptoms or ADHD being bullied; providing targeted support for more vulnerable children may also form an important component of future intervention efforts.

Conflicts of interest

Lucy Bowes receives funding from the Academy of Medical Sciences and the NIHR for her research in bullying.

Links

Primary paper

Schoeler T, Choi SW, Dudbridge F, et al. (2019) Multi–Polygenic Score Approach to Identifying Individual Vulnerabilities Associated With the Risk of Exposure to Bullying (PDF). JAMA Psychiatry. Published online April 03, 2019. doi:10.1001/jamapsychiatry.2019.0310

Other references

Arseneault, L., L. Bowes and S. Shakoor (2010). “Bullying victimization in youths and mental health problems: ‘much ado about nothing’?” Psychol Med 40(5): 717-729. https://doi.org/10.1017/S0033291709991383

Ball, H. A., L. Arseneault, A. Taylor, B. Maughan, A. Caspi and T. E. Moffitt (2008). “Genetic and environmental influences on victims, bullies and bully-victims in childhood.” J Child Psychol Psychiatry 49(1): 104-112. https://doi.org/10.1111/j.1469-7610.2007.01821.x

Bowes, L., L. Arseneault, B. Maughan, A. Taylor, A. Caspi and T. E. Moffitt (2009). “School, neighborhood and family factors are associated with children’s bullying involvement: A nationally-representative longitudinal study.” Journal of the American Academy of Child and Adolescent Psychiatry 48: 545-553. https://doi.org/10.1097/CHI.0b013e31819cb017

Brugger, S. (2017). Bullying in childhood: cause or consequence of mental health problems? #AntiBullyingWeek. The Mental Elf, 2017.

Copeland, W. E., D. Wolke, A. Angold and E. Costello (2013). “Adult psychiatric outcomes of bullying and being bullied by peers in childhood and adolescence.” JAMA Psychiatry 70(4): 419-426. https://doi.org/10.1001/jamapsychiatry.2013.504

Ford, T. (2018) Bullying in childhood and adolescence: we need to do better. The Mental Elf, 2018.

Przybylski, A. K. and L. Bowes (2017). “Cyberbullying and adolescent well-being in England: a population-based cross-sectional study.” Lancet Child Adolesc Health 1(1): 19-26. https://doi.org/10.1016/S2352-4642(17)30011-1

Reynolds S. (2017) Cyberbullying: comparatively rare, not especially damaging or pernicious. The Mental Elf, 2017.

Ttofi, M. M. and D. P. Farrington (2011). “Effectiveness of school-based programs to reduce bullying: a systematic and meta-analytic review.” J Exp Criminol 7: 27-56. https://doi.org/10.1007/s11292-010-9109-1

Wertz, J. (2016). Childhood bullying and mental illness in young adulthood. The Mental Elf, 2016.

Wolke, D., W. E. Copeland, A. Angold and E. J. Costello (2013). “Impact of bullying in childhood on adult health, wealth, crime, and social outcomes.” Psychol Sci 24(10): 1958-1970. https://doi.org/10.1177/0956797613481608