Examining the effect of smoking on suicidal ideation and attempts: A triangulation of epidemiological approaches

Background Previous literature has demonstrated a strong association between cigarette smoking and suicide-related behaviours, characterised as ideation, plans, attempts and suicide related death. This association has not previously been examined in a causal inference framework and has important implications for suicide prevention strategies. Aims We aimed to examine the evidence for an association between smoking behaviours (initiation, smoking status, heaviness, lifetime smoking) and suicidal thoughts or attempts by triangulating across observational and Mendelian randomisation (MR) analyses. Methods First, in the UK Biobank, we calculate observed associations between smoking behaviours and suicidal thoughts or attempts. Second, we used Mendelian randomisation (MR) to explore the relationship between smoking and suicide using genetic variants as instruments to reduce bias from residual confounding and reverse causation. Results Our observational analysis showed a relationship between smoking behaviour and suicidal behaviour, particularly between smoking initiation and suicidal attempts (OR = 2.07, 95% CI = 1.91 to 2.26, p<0.001). The MR analysis and single SNP analysis, however, did not support this. Despite past literature showing a positive dose-response relationship our results showed no clear evidence for a causal effect of smoking on suicidal behaviours. Conclusion This was the first MR study to explore the effect of smoking on suicidal behaviours. Our results suggest that, despite observed associations, there is no strong evidence for a causal effect of smoking behaviour on suicidal behaviour. Our evidence suggests that further research is needed into alternative risk factors for suicide which might make better intervention targets.