Mendelian randomisation studies of Attention Deficit Hyperactivity Disorder.

Background: Observational studies have found Attention Deficit Hyperactivity Disorder (ADHD) to be associated with an increased risk of adverse outcomes as well as with early risk factors; however it is not clear whether these associations reflect causal effects. Alternatives to traditional observational studies are needed to investigate causality: one such design is Mendelian randomization (MR), which uses genetic variants as instrumental variables for the exposure. Methods: In this review we summarise findings from approximately 50 studies using MR to examine potentially causal associations with ADHD as either an exposure or outcome. Results: To-date, few MR ADHD studies have investigated causal evidence with other neurodevelopmental, mental health and neurodegenerative conditions but those that have suggest a complex relationship with autism, some evidence of a causal effect on depression and limited evidence of a causal effect on neurodegenerative conditions. For substance use, MR studies provide evidence consistent with a causal effect of ADHD on smoking initiation, but findings for other smoking behaviours and cannabis use are less consistent. Studies of physical health suggest bidirectional causal effects with higher body mass index, with stronger effects for childhood obesity, as well as some evidence of causal effects on coronary artery disease and stroke in adults and limited evidence of causal effects on other physical health problems or sleep. Studies suggest bidirectional relationships between ADHD and socio-economic markers and provide some evidence that low birthweight may be a causal risk factor for ADHD, while bidirectional evidence has been found for some environmental factors. Finally, there is emerging evidence of bidirectional causal links between ADHD genetic liability and biological markers of human metabolism and inflammation. Conclusions: While MR has advantages over traditional observational designs in addressing causality, we discuss limitations of current ADHD studies and future directions, including the need for larger genome-wide association studies (and using samples of different ancestries), and for triangulation with different methods.