Background/Purpose: In observational studies, serum urate is positively associated with bone mineral density (BMD) and reduced risk of fragility fractures. However, the possibility of unmeasured confounding means that it is uncertain whether urate is a direct mediator of bone density.  Mendelian randomization analysis assumes that inherited genetic risk variants for one phenotype are naturally randomized with respect to a second unrelated phenotype, and allows disentangling of cause and effect in the presence of potential confounding. We used Mendelian randomization analysis to examine whether there is a causal relationship between serum urate and BMD.

Methods: We analysed data from the Generation 3 cohort in the Framingham Heart Study (FHS).  Inclusion criteria were availability of serum urate, BMD, body mass index (BMI) and genotype data.  Exclusion criteria were eGFR<30, and any of the following medications: diuretics, bisphosphonates, oral glucocorticoids and hormone replacement therapy.  A weighted urate genetic risk score (GRS) was calculated using SNPs for SLC2A9, ABCG2, SLC17A1, SLC22A11 and SLC22A12 based on the genome wide association study by Kottgen et al¹. Mendelian randomization analysis was done using the two-stage least squares (2SLS) method, this first calculates each individual’s genetically determined predictor levels (predicts urate levels based on the urate GRS) then regresses these calculated predictor values against the outcome (BMD). The analysis was adjusted for age, sex, menopause status in women, BMI and the first two principal components from genome-wide genotype data.

Results: There were 2553 eligible participants.  The urate GRS accounted for 3.3% of the variance in serum urate in the total group.  A strong association between serum urate at Examination 1 and BMD at Examination 2 was observed in the unadjusted ordinary least square analysis (p<8E-17 for all BMD measures, Table), which was attenuated but persisted at the spine and total femur following adjustment for potential confounders.  In the genetic 2SLS analysis, no significant relationship was observed between urate and BMD, either in the unadjusted or adjusted analysis.  For total femur and femoral neck BMD, the effect sizes for the ordinary least square and 2SLS analysis were significantly different (Durbin-Hausman p<0.02).  Similar findings were observed in both the male and female subgroups.  Analysis of a causal role for BMD on serum urate levels using a weighted BMD GRS did not demonstrate evidence for reverse causation (data not shown).  

Conclusion: Serum urate is strongly associated with BMD.  However, controlling for confounders by Mendelian randomization does not provide evidence that serum urate is causal in increased BMD.  

Reference: Kottgen et al Nature Genetics 2013.

Table: Mendelian randomization analysis with BMD as the outcome of interest for all eligible participants.