Long-term triglyceride glucose index (TyG)-uric acid burden and incident stroke in the China Health and Retirement Longitudinal Study (CHARLS): repeated-exposure and transition analyses

The uric acid index (UAI) combines the triglyceride-glucose index (TyG) with log-transformed uric acid as a composite measure of metabolic–urate exposure. The incremental information provided by repeated UAI measurements beyond TyG remains uncertain. We examined two-examination mean UAI and observed hypertension–stroke state sequences in the China Health and Retirement Longitudinal Study (CHARLS). We studied 4,140 adults with fasting UAI measurements in 2011 and 2015, no recorded cardiovascular disease through 2015, and subsequent outcome observation. UAI equals TyG plus ln(uric acid); its arithmetic mean across the two examinations defined the repeated exposure. Community-clustered complementary log-log models assessed self-reported physician-diagnosed stroke in 2018–2020. A separate 2011 cohort without baseline cardiovascular disease or hypertension supported analyses of recorded state sequences. We compared components and adjustment strategies on identical samples and examined selection, recorded state sequences, and incremental prediction. There were 291 strokes; the fully adjusted analysis included 3,503 participants and 253 events. Per one reference standard deviation (0.659 units) higher mean UAI, the hazard ratio (HR) was 1.21 (95% confidence interval [CI] 1.03–1.42). In 3,358 participants with both covariate sets, the HR was 1.27 (1.08–1.50) with 2011 adjustment and 1.14 (0.96–1.35) with 2015 adjustment. Mean TyG was associated with stroke (HR 1.26, 1.08–1.47); adding ln(uric acid) provided no detectable incremental association (cluster-robust P = 0.353). Baseline UAI was associated with stroke first recorded without preceding recorded hypertension (HR 1.31, 1.13–1.51), an observed sequence that does not establish biological ordering. No subgroup interaction survived correction for six tests. Adding mean UAI did not demonstrate improved optimism-corrected discrimination over clinical variables alone (AUC 0.671 vs. 0.672). Two-examination mean UAI was associated with incident stroke, but the estimate was sensitive to covariate timing and selection assumptions. These data do not demonstrate incremental value of its uric acid component beyond TyG or support clinical use of UAI for stroke prediction. The state-sequence analyses are descriptive and do not establish or exclude mediation by hypertension.