Mediation analysis seeks to explain the pathway(s) through which an exposure affects an outcome. Traditional, non-instrumental variable methods for mediation analysis experience a number of methodological difficulties, including bias due to confounding between an exposure, mediator and outcome and measurement error. Mendelian randomisation (MR) can be used to improve causal inference for mediation analysis. We describe two approaches that can be used for estimating mediation analysis with MR: multivariable MR (MVMR) and two-step MR. We outline the approaches and provide code to demonstrate how they can be used in mediation analysis. We review issues that can affect analyses, including confounding, measurement error, weak instrument bias, interactions between exposures and mediators and analysis of multiple mediators. Description of the methods is supplemented by simulated and real data examples. Although MR relies on large sample sizes and strong assumptions, such as having strong instruments and no horizontally pleiotropic pathways, our simulations demonstrate that these methods are unaffected by confounders of the exposure or mediator and the outcome and non-differential measurement error of the exposure or mediator. Both MVMR and two-step MR can be implemented in both individual-level MR and summary data MR. MR mediation methods require different assumptions to be made, compared with non-instrumental variable mediation methods. Where these assumptions are more plausible, MR can be used to improve causal inference in mediation analysis.

The goal of much observational research is to identify risk factors that have a causal effect on health and social outcomes. However, observational data are subject to biases from confounding, selection and measurement, which can result in an underestimate or overestimate of the effect of interest. Various advanced statistical approaches exist that offer certain advantages in terms of addressing these potential biases. However, although these statistical approaches have different underlying statistical assumptions, in practice they cannot always completely remove key sources of bias; therefore, using design-based approaches to improve causal inference is also important. Here it is the design of the study that addresses the problem of potential bias – either by ensuring it is not present (under certain assumptions) or by comparing results across methods with different sources and direction of potential bias. The distinction between statistical and design-based approaches is not an absolute one, but it provides a framework for triangulation – the thoughtful application of multiple approaches (e.g. statistical and design based), each with their own strengths and weaknesses, and in particular sources and directions of bias. It is unlikely that any single method can provide a definite answer to a causal question, but the triangulation of evidence provided by different approaches can provide a stronger basis for causal inference. Triangulation can be considered part of wider efforts to improve the transparency and robustness of scientific research, and the wider scientific infrastructure and system of incentives.

Depression is a leading cause of disability worldwide¹ and is expected to be the highest global burden of disease by 2030.² Despite efforts to improve interventions, prevalence is still increasing, especially in adolescence.³ Evidence suggests that depression during adolescence is associated with many concurrent and later psychological and social impairments.^4,5 However, what is driving this increase in adolescent depression is still not clear. A greater understanding of the nature of adolescent depression and how to minimize it is crucial if we are to reduce this global burden.

The estimated worldwide prevalence of non-alcoholic fatty liver disease (NAFLD) in adults is 25%; however, prevalence in young adults remains unclear. We aimed to identify the prevalence of steatosis and fibrosis in young adults in a sample of participants recruited through the Avon Longitudinal Study of Parents and Children (ALSPAC), based on transient elastography and controlled attenuation parameter (CAP) score. In this population-based study, we invited active participants of the ALSPAC cohort to our Focus@24+ clinic at the University of Bristol (Bristol, UK) between June 5, 2015, and Oct 31, 2017, for assessment by transient elastography with FibroScan, to determine the prevalence of steatosis and fibrosis. FibroScan data were collected on histologically equivalent fibrosis stage (F0-F4) and steatosis grade (S0-S3); results with an IQR to median ratio of 30% or greater were excluded for median fibrosis results greater than 7·1 kPa, and CAP scores for steatosis were excluded if less than ten valid readings could be obtained. Results were collated with data on serology (including alanine aminotransferase, aspartate aminotransferase, and γ-glutamyl transferase) and exposures of interest: alcohol consumption (via the Alcohol Use Disorder Identification Test for Consumption [AUDIT-C] and the Diagnostic and Statistical Manual of Mental Disorders-5 criteria for alcohol use disorder), body-mass index (BMI), waist-to-height ratio, socioeconomic status (based on predefined ALSPAC markers), and sex. We used logistic regression models to calculate odds ratios (ORs) for the effect of exposures of interest on risk of steatosis and fibrosis, after dichotomising the prevalences of fibrosis and steatosis and adjusting for covariates (excessive alcohol intake [hazardous drinking, AUDIT-C score ≥5; or harmful drinking, evidence of alcohol use disorder], social class, smoking, and BMI). 10 018 active ALSPAC participants were invited to our Focus@24+ clinic, and 4021 attended (1507 men and 2514 women), with a mean age of 24·0 years (IQR 23·0-25·0). 3768 CAP scores were eligible for analysis. 780 (20·7% [95% CI 19·4-22·0]) participants had suspected steatosis (S1-S3; ≥248 dB/m), with 377 (10·0%) presenting with S3 (severe) steatosis (≥280 dB/m). A BMI in the overweight or obese range was positively associated with steatosis when adjusted for excessive alcohol consumption, social class, and smoking (overweight BMI: OR 5·17 [95% CI 4·11-6·50], p<0·0001; obese BMI: 27·27 [20·54-36·19], p<0·0001). 3600 participants had valid transient elastography results for fibrosis analysis. 96 participants (2·7% [95% CI 2·2-3·2]) had transient elastography values equivalent to suspected fibrosis (F2-F4; ≥7·9 kPa), nine of whom had values equivalent to F4 fibrosis (≥11·7 kPa). Individuals with alcohol use disorder and steatosis had an increased risk of fibrosis when adjusted for smoking and social class (4·02 [1·24-13·02]; p=0·02). One in five young people had steatosis and one in 40 had fibrosis around the age of 24 years. The risk of fibrosis appears to be greatest in young adults who have harmful drinking patterns and steatosis. A holistic approach to the UK obesity epidemic and excessive drinking patterns is required to prevent an increasing health-care burden of adults with advanced liver disease in later life. Medical Research Council UK, Alcohol Change UK, David Telling Charitable Trust.

Young people whose parents have depression have a greatly increased risk of developing a psychiatric disorder, but poor outcomes are not inevitable. Identification of the contributors to mental health resilience in young people at high familial risk is an internationally recognised priority. Our objectives were to identify protective factors that predict sustained good mental health in adolescents with a parent with depression and to test whether these contribute beyond what is explained by parent illness severity. The Early Prediction of Adolescent Depression study (EPAD) is a prospective longitudinal study of offspring of parents with recurrent depression. Parents with recurrent major depressive disorder, co-parents, and offspring (aged 9-17 years at baseline) were assessed three times over 4 years in a community setting. Offspring outcomes were operationalised as absence of mental health disorder, subthreshold symptoms, or suicidality on all three study occasions (sustained good mental health); and better than expected mental health (mood and behavioural symptoms at follow-up lower than predicted given severity of parental depression). Family, social, cognitive, and health behaviour predictor variables were assessed using interview and questionnaire measures. Between February and June, 2007, we screened 337 families at baseline, of which 331 were eligible. Of these, 262 completed the three assessments and were included in the data for sustained mental health. Adolescent mental health problems were common, but 53 (20%) of the 262 adolescents showed sustained good mental health. Index parent positive expressed emotion (odds ratio 1·91 [95% CI 1·31-2·79]; p=0·001), co-parent support (1·90 [1·38-2·62]; p<0·0001), good-quality social relationships (2·07 [1·35-3·18]; p=0·001), self-efficacy (1·49 [1·05-2·11]; p=0·03), and frequent exercise (2·96 [1·26-6·92]; p=0·01) were associated with sustained good mental health. Analyses accounting for parent depression severity were consistent, but frequent exercise only predicted better than expected mood-related mental health (β=-0·22; p=0·0004) not behavioural mental health, whereas index parents' expression of positive emotions predicted better than expected behavioural mental health (β=-0·16; p=0·01) not mood-related mental health. Multiple protective factors were required for offspring to be free of mental health problems (zero or one protective factor, 4% sustained good mental health; two protective factors, 10%; three protective factors, 13%, four protective factors, 38%; five protective factors, 48%). Adolescent mental health problems are common, but not inevitable, even when parental depression is severe and recurrent. These findings suggest that prevention programmes will need to enhance multiple protective factors across different domains of functioning. Sir Jules Thorn Charitable Trust, Economic and Social Research Council.

There is increasing evidence that childhood Attention-Deficit Hyperactivity Disorder (ADHD) elevates risk of later depression, but the mechanisms behind this association are unclear. We investigated the relationship between childhood ADHD symptoms and late-adolescent depressive symptoms in a population cohort, and examined whether academic attainment and peer problems mediated this association. ALSPAC (Avon Longitudinal Study of Parents and Children) is an ongoing prospective longitudinal population-based UK cohort that has collected data since September 1990. 2950 individuals with data on parent-reported ADHD symptoms in childhood (7.5 years) and self-reported depressive symptoms in late adolescence (17.5 years) were included in analyses. 2161 individuals with additional data at age 16 years on parent-reported peer problems as an indicator of peer relationships and formal examination results (General Certificate of Secondary Education; GCSE) as an indicator of academic attainment were included in mediation analyses. Childhood ADHD symptoms were associated with higher depressive symptoms (b = 0.49, SE = 0.11, p < 0.001) and an increased odds of clinically significant depressive symptoms in adolescence (OR = 1.27, 95% CI 1.15–1.41, p < 0.001). The association with depressive symptoms was mediated in part by peer problems and academic attainment which accounted for 14.68% and 20.13% of the total effect, respectively. Childhood ADHD is associated with increased risk of later depression. The relationship is mediated in part by peer relationships and academic attainment. This highlights peer relationships and academic attainment as potential targets of depression prevention and intervention in those with ADHD. Future research should investigate which aspects of peer relationships are important in conferring later risk for depression.