Background Smoking is strongly associated with lower body mass index (BMI) and smokers cite loss of weight control as a powerful disincentive to quit smoking. However, it is not clear which aspects of weight distribution are affected by smoking, and observational evidence suggests waist circumferences and waist-hip ratio may be higher in smokers than in non-smokers after adjusting for BMI.

Methods We performed a Mendelian randomisation analysis, using data on over 139,000 individuals from 29 studies in the Consortium for Causal Analysis Research in Tobacco and Alcohol (CARTA) to investigate the causal nature of associations of smoking with a range of regional adiposity phenotypes. We used a smoking related variant (rs1051730/rs16969968) for which each copy of the minor allele is associated with increasing smoking consumption by approximately one cigarette per day. We sought per allele associations of this variant with BMI, weight, waist circumference, hip circumference and waist-hip ratio (all log transformed), using linear regression, adjusted for age and stratified by sex and smoking status (never, former and current). All analyses were additionally adjusted for BMI. Results from all studies were combined in meta-analyses. Interactions between smoking status and genotype, and sex and genotype, were assessed using the Cochran Q statistic.

Results Each additional minor allele copy of the smoking related variant was associated with decreases in adiposity measures in current smokers: per allele percentage change in BMI: -0.74% (95% CI -0.96, -0.51; p < 0.001); in waist circumference -0.37% (-0.55, -0.19; p < 0.001); in hip circumference -0.31% (-0.42, -0.19, p < 0.001) and in waist-hip ratio -0.08% (-0.19, 0.03). These per allele changes differed significantly from those seen among ex-smokers or never-smokers in whom changes were more modest; for example changes in BMI were -0.14% (-0.34, 0.07) among ex-smokers and 0.35% (0.18, 0.52) among never-smokers: these effects differed highly significantly among smoking status groups, but very little between studies within these groups. After adjustment for BMI, per allele changes among current smokers became positive for waist circumference: 0.14% (0.05, 0.23), and waist-hip ratio: 0.10% (0.02, 0.19) while no clear relationship could be seen for hip circumference 0.04% (-0.01, 0.09).

Conclusion A variant associated with increased smoking consumption was also associated with lower BMI among current smokers, as well as lower waist and hip circumference and waist-hip ratio, albeit of lesser magnitude than for BMI. After adjustment for BMI, the association for waist circumference was positive but unclear for hip circumference. This suggests that higher cigarette consumption may lead to a preferential re-distribution towards central adiposity.