Are blood lipids risk factors for fracture? Integrative evidence from instrumental variable causal inference and mediation analysis using genetic data.

Background: The relationship between lipids and the risk of fracture is currently controversial and whether such association is causal remains elusive.

Methods: We performed two-sample inverse variance weighted (IVW) Mendelian randomization (MR) analyses to evaluate causal effects of four lipids (i.e. high-density lipoprotein cholesterol [HDL], low-density lipoprotein cholesterol [LDL], total cholesterol [TC] and triglyceride [TG]) on fracture or bone mineral density (BMD) with summary statistics from large scale genome-wide association studies (up to ~190,000 for lipids, ~66,628 for BMD and ~53,000 for fracture). We validated our MR results with extensive sensitive analyses including MR-PRESSO and MR-Egger regression. Multivariable analyses were implemented to investigate whether other lipids (i.e. LDL and TG) may confound the causal effect of HDL on fracture and mediation analyses were conducted to assess indirect effects of lipids on fracture mediated by BMD.

Results: The IVW MR showed there existed a statistically significant association between HDL and fracture, with the odd ratio (OR) per standard deviation change of HDL on fracture being 1.12 (95% CI: 1.02-1.22, p = 1.20E-02). HDL was also detected to be causally associated with BMD (beta = -0.116; 95% CI: -0.182 ~ -0.050, p = 5.47E-04). These associations were further confirmed by the weighted median and maximum likelihood methods, with the MR-Egger regression removing the possibility of pleiotropy and the multivariable analysis excluding the confounding effect of other lipids on HDL. Negative associations of HDL with BMD among the elderly and with BMD at the lumbar spine were also discovered. However, no causal associations were detected between other lipids (OR = 0.87, 95% CI: 0.74-1.03, p = .107 for LDL; OR = 1.03; 95% CI: 0.88-1.21, p = .696 for TC and OR = 1.04; 95% CI: 0.90-1.20, p = .610 for TG) and fracture; whereas TG was positively associated BMD (beta = 0.184; 95% CI: 0.048-0.319, p = 7.93E-03). Finally, the mediation effect of BMD was estimated to be -0.116 (95% CI: -0.182 to -0.05, p = 5.47E-04) for HDL or 0.184 (95% CI: 0.048-0.319, p = 7.93E-03) for TG, implying HDL and TG could be indirectly associated with fracture risk via the pathway of BMD.

Conclusion: Our study is supportive of the causal relationship between HDL and fracture but offers little direct evidence for causal associations between other lipids and fracture, and further reveals HDL and TG may have an indirect influence on fracture mediated by BMD.