Objectives: Despite the inverse association between cigarette smoking and body mass index (BMI), it is unknown whether the effect of smoking on insulin is mediated through decreased BMI. This study aims to examine the temporal relationship between BMI and insulin, the impact of smoking on this relationship and the mediation effect of BMI on the association between smoking and insulin levels.
Methods: The longitudinal cohort consisted of 1121 adults (807 white and 314 black participants, mean age, 42.0 years at follow-up) for whom BMI and fasting insulin were measured twice, with an average follow-up period of 17.1 years. Cross-lagged panel and mediation analysis models were used to examine the temporal relationship between BMI and insulin, and the mediation effect of BMI on the smoking-insulin association.
Results: Smoking was inversely associated with insulin (regression coefficient, -0.073; P = .015 at baseline and -0.121; P < .001 at follow-up), adjusting for age, race and gender. After additional adjustment for follow-up periods, the cross-lagged path coefficient from BMI to insulin (β, 0.226; P < .001) was significantly greater than that from insulin to BMI (β, -0.029; P = .208), with P < .001 for the difference. The path coefficient from BMI to insulin was significantly greater in non-smokers (β, 0.273; P < .001) than in smokers (β, 0.122; P = .046), with P = .013 for the difference. The mediation effect of BMI on the smoking-insulin association was estimated at 53.4% (P = .030) at baseline and 58.7% (P < .001) at follow-up.
Conclusions: These findings suggest that cigarette smoking has a significant impact on the one-directional relationship from BMI to insulin. The insulin-lowering effect of smoking is predominantly mediated through decreased BMI as the result of smoking.
Keywords: body mass index; insulin; mediation; smoking; temporal relationship.
© 2018 John Wiley & Sons Ltd.