Background/Purpose: Obesity is a strong risk factor of incident gout, but previous research showed no such association with recurrent gout among gout patients. These paradoxical findings occur because the causal net effect (i.e., total effect) of obesity on recurrent gout in gout patients cannot be validly estimated using conventional methods. We demonstrate that the paradox can be clarified using an appropriate mediation analysis, and illustrate a design to estimate the total effect of BMI on recurrent gout in incident gout patients.

Methods: We used data from the Multiple Risk Factor Intervention Trial (MRFIT), prospectively collected at baseline and annually over 7 years.  BMI at baseline was categorized as: >30 kg/m²(obese), 25-29.9 (overweight), and <25 (normal) and self-report of physician-diagnosis of gout was our outcome. We followed subjects without gout at baseline to determine the first occurrence of gout and their recurrent gouts by the 84-month visit. We assessed the effect of BMI on recurrent gout with the conventional method of restricting on incident gout patients and using logistic regression. We then clarified the paradox using marginal structural modeling (MSM) for mediation analysis. We estimated the total effect of BMI on recurrent gout by decomposing the total effect into its components (Figure): the indirect effect of BMI via its effect on incident gout, and the direct effect of BMI on recurrent gout not through its effect on incident gout.  Finally, we determined the association of change in BMI categories before and after incident gout on risk of recurrent gout among incident gout patients. All analyses were adjusted for known confounders. 

Results: Of 11,655 subjects without gout at baseline (mean age 46 years; 21% normal, 56% overweight, 23% obesity), 408 people developed incident gout, and 131 had recurrent gout.  Conventional method showed that the adjusted odds ratio (OR) for recurrent gout was 1.10 (95%CI: 0.52,2.30) for obese compared with normal BMI (Table). Using MSM, the indirect effect of obesity compared with normal-weight on risk of recurrent gout (via its effect on incident gout) was 2.62 (95%CI:2.01,3.40); the direct effect not through incident gout was 1.13 (95%CI:0.82,1.56); and the total effect was 2.94 (95%CI:1.59,5.41) (Table).  Among incident gout patients, the adjusted total effect of increasing BMI after incident gout on risk of recurrent gout was 1.84 (95%CI:1.02,3.32) (Table).

Conclusion:  We showed that the effect of obesity at baseline on risk of recurrent gout is almost entirely through its effect on incident gout.  Conditioning on incident gout and estimating the effect of baseline obesity on the risk of recurrent gout would provide only an estimate of its direct effect, as the indirect effect is blocked.  In order to examine the total effect of obesity on recurrent gout in those with incident gout, BMI change must be assessed before and after incident gout.