The role of familial confounding in the associations of physical activity, smoking and alcohol consumption with early exit from the labour market.

We investigated the associations between health behaviors and sustainable working life outcomes including all-cause disability pension, disability pensions due to musculoskeletal and mental diagnoses and unemployment. The role of familial factors behind these associations was studied by analysing discordant twin pairs. Our data included Swedish twins born in 1925-1986 (51891 twin individuals). Baseline data based on two independent surveys in 1998-2003 and 2005-2006 for health behaviors were linked to national registers on disability pension and unemployment until 2016. Cox proportional hazards models for hazard ratios (HR) with 95% confidence intervals (CI) were estimated for the whole sample adjusting for covariates. Analyses of health behavior discordant twin pairs (n = 5903 pairs) were conducted using conditional Cox models. In the whole cohort, the combination of healthy behaviors was associated with lower risk for all-cause disability pension, disability pension due to musculoskeletal diagnoses or mental diagnoses, and for unemployment (HRs 0.56-0.86, 95% CIs 0.51-0.92) as did being physically active (HRs 0.69-0.87, 95% CI 0.65-0.92). The discordant pair analyses confirmed the lower risk among those having healthy behaviors (HR 0.70-0.86) or being physically active (HR 0.86-0.87) for all-cause disability pension, disability pension due to musculoskeletal diagnoses, and for unemployment. To conclude, controlling the effects of covariates or familial confounding (i.e. discordant twin pair analyses) shows that being physically active or having several healthy behaviors predict better working life outcomes. This points towards independent association between healthy behavior and longer working life.