A multivariable Mendelian randomization study of serum lipids and dementia risk within the UK Biobank

Background An unfavourable lipid profile has been associated with increased risk of dementia. However, it is challenging to investigate each serum lipid measure individually due to the high correlation between the traits. We tested for genetic evidence supporting associations between serum lipid measures and risk of dementia. Methods We conducted multivariable and univariable Mendelian randomisation (MR) analyses on 329,896 UK Biobank participants (age 37–73 years) to examine the associations between low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglycerides, apolipoprotein-A1 (ApoA1) and apolipoprotein-B (ApoB), and the risk of dementia. The multivariable approach allows us to assess the association of each lipid measure with the outcome, including where the genetic variant-exposure associations are mediated by one another. Results In the univariable MR analyses, we observed no association between genetically determined serum lipids and risk of dementia. However, in a multivariable MR model containing LDL-C, triglycerides, and ApoB, ApoB was associated with a higher risk of dementia (OR per 1 SD higher ApoB 1.63, 95% CI 1.12, 2.37). Multivariable findings were consistent across IVWMR and MR-Egger, but not weighted median MR or MR-Lasso. HDL-C and ApoA1 were not associated with dementia in univariable or multivariable MR. Conclusions Our findings suggest that when considering the correlation between lipid measures, ApoB may play a role in the previously reported association between serum lipids and increased risk of dementia. Future studies should aim to confirm the findings in clinical/experimental studies and further explore the role of ApoB in dementia pathophysiology.