Abstract
AIM: A Mendelian randomization study (MRS) can be linked to a "natural" randomized controlled trial in order to avoid potential bias of observational epidemiology. We aimed to study the possible association between serum urate (SU) and total bilirubin (TBIL) using MRS. MATERIALS AND METHODS: An observational epidemiological study using ordinary least squares (OLS) regression and MRS using two-stage least square (TLS) regression was conducted to assess the effect of SU on TBIL. The comparison between the OLS regression and the TLS regression was analyzed by the Durbin-Hausman test. If the p value is significant, it suggests that the OLS regression cannot evaluate the relationship between exposure and outcome, and the TLS regression is precise; while if the p value is not significant, there would be no significant difference between the two regressions. RESULTS: A total of 3,753 subjects were analyzed. In OLS regression, there was no significant association between SU and TBIL in all subjects and subgroup analysis (all p > 0.05). However, MRS revealed a negative correlation between SU and TBIL after adjustment for confounders (beta = -0.021, p = 0.010). Further analysis was conducted in different SU subgroups, and results show that elevated SU was associated with a significant reduction in TBIL after adjustment for hyperuricemic subjects (beta = -0.053, p = 0.027). In addition, the results using the Durbin-Hausman test further confirmed a negative effect of SU on TBIL (p = 0.002 and 0.010, respectively). CONCLUSIONS: This research shows for the first time that elevated SU was a potential causal factor in the reduction of TBIL and it provides strong evidence to resolve the controversial association between SU and TBIL.