Association between adiposity distribution and low-grade systemic inflammation: Tackling multicollinearity

To examine the association between adiposity distribution and low-grade systemic inflammation, testing how the statistical approach can be used to tackle the multicollinearity between variables of adiposity.

Cross-sectional analysis of 987 adults (27 years old) from the EPITeen cohort (Porto, Portugal). Adiposity distribution was assessed based on the fat distribution of the participants, measured by whole-body dual-energy X-ray absorptiometry (DXA). Low-grade systemic inflammation was determined based on the serum high-sensitivity C-reactive protein (hsCRP) of the participants. The associations were estimated by linear regression models, stratified by sex, and multicollinearity between variables of adiposity was considered.

The association (standardized β [95%CI]) between total body fat and hsCRP (ln) was 0.332 (0.252, 0.411) in women and 0.256 (0.175, 0.337) in men. Finally, in women, the association (standardized β [95%CI]) with hsCRP (ln) was -0.014 (-0.094, 0.065) for limb fat and 0.019 (-0.061, 0.099) for trunk fat, and, in men it was -0.136 (-0.216, -0.056) for limb fat and 0.130 (0.050, 0.211) for trunk fat.

After a comprehensive study of fat, its overall proinflammatory effect was acknowledged. Trunk fat was more proinflammatory than limb fat, even though the results were only statistically significant in men. Statistical analysis should consider multicollinearity between variables of adiposity, though simpler models could be used to address this problem.