Endocrine Abstracts

Background: We used Mendelian randomization and machine learning (gradient boosting) to assess the causal association of Type 2 Diabetes with osteoporosis.

Methods: We selected 155 SNPs associated with type 2 diabetes and glycaemic and insulin-related traits reported in previous studies (1,2) and studied the association with Heel bone mineral density (BMD)) (n= 4 72 996) and risk of fragility fractures (26 157 cases and 17 807 controls) in UKBiobank participants. The association of the SNPs was tested with BMD using a linear regression model with adjustment for top 5 principal components, age, sex, smoking and BMI. The association of the top SNPs (P<10-8) was also tested in a subset sample of 26 157 cases with fragility fractures and 17 807 controls using a logistic regression model with covariate adjustments. In the gradient boosting analysis those with BMD less than the first quartile of distribution was categorised as low BMD while those above it as high BMD and age, sex, smoking and BMI and 155 SNPs were included as predictor variables. All the statistical analysis was done in R3.5.5.

Results: The study consisted of 4 72 996 participants in UKBiobank 52% female with median BMD of 0.52 (IQR 0.44–0.60). Of the 155 SNPs associated with T2D and glycaemic traits, 17 SNPs from 11 gene/loci were associated with BMD at genome-wide significance (P<1-08). The top associated SNPs with BMD included rs2745353 in RSPO3, rs983309 in RP11, rs6072275 in RP1 rs174576 in FADS2, rs10203174 in THADA and rs1727313 in MPHOSPH9. SNPs in RSPO3 and RP11 were associated with the risk of fragility fractures. The gradient boosting analysis confirmed the results of the regression analysis.

Conclusion: We used Mendelian randomization and machine learning to show that type 2 diabetes is causally associated with BMD and risk of fragility fractures.