Premium
Obesity Paradox in Recurrent Attacks of Gout in Observational Studies: Clarification and Remedy
Author(s) -
Nguyen UyenSa D. T.,
Zhang Yuqing,
LouieGao Qiong,
Niu Jingbo,
Felson David T.,
LaValley Michael P.,
Choi Hyon K.
Publication year - 2017
Publication title -
arthritis care and research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.032
H-Index - 163
eISSN - 2151-4658
pISSN - 2151-464X
DOI - 10.1002/acr.22954
Subject(s) - gout , medicine , body mass index , obesity , odds ratio , risk factor , physical therapy , surgery
Objective Obesity is strongly associated with incident gout risk, but its association with the risk of recurrent gout attacks has been null or weak, constituting an obesity paradox. We sought to demonstrate and overcome the methodologic issues associated with the obesity paradox for risk of recurrent gout attacks. Methods Using the Multiple Risk Factor Intervention Trial database, we decomposed the total effect of obesity into its direct and indirect (i.e., mediated) effects using marginal structural models. We also estimated the total effect of body mass index (BMI) change from baseline among incident gout patients. Results Of 11,816 gout‐free subjects at baseline, we documented 408 incident gout cases, with 132 developing recurrent gout attacks over a 7‐year followup period. The adjusted odds ratio (OR) for incident gout among obese individuals was 2.6, whereas that for recurrent gout attacks among gout patients was 0.98 (i.e., the obesity paradox). These ORs correlated well with the ORs for the indirect and direct effects of obesity on the risk of recurrent gout attacks (2.83 and 0.98, respectively). Compared with no BMI change, the OR of increasing versus decreasing >5% of baseline BMI was 0.61 and 1.60 for recurrent gout attacks, respectively ( P for trend < 0.01), suggesting a dose‐response association. Conclusion The obesity paradox for the risk of recurrent gout attacks is explained by the absence of the direct effect, which is often measured in conventional analyses and misinterpreted as the intended total effect of interest. In contrast, the BMI change analysis correctly estimated the intended total effect of BMI, and revealed a dose‐response relationship.