Longitudinal monitoring of uric acid-to-eGFR ratio for CKD prediction in community older adults with unstable renal function: A prospective cohort study.

BACKGROUND

Older adults with fluctuating renal function face elevated chronic kidney disease (CKD) risk, yet relevant predictive indicators are lacking. We hypothesized that monitoring of the two-year average uric acid-to-eGFR ratio (aUFR) could dynamically predict incident CKD in this population.

METHODS

This prospective cohort study included 2651 community-dwelling older adults with transient kidney function fluctuation, defined as an isolated reduction in eGFR during the initial assessment period. Cox proportional hazards models, longitudinal trajectory analyses, and restricted cubic spline models were used to evaluate the association between aUFR and incident CKD, whereas ROC analysis assessed its discriminative performance. Subgroup and sensitivity analyses examined the robustness of the findings.

RESULTS

During the 3-year follow-up, 613 participants (23.12%) developed CKD. Each 1-unit increase in aUFR was independently associated with a 67% higher risk of incident CKD (fully adjusted HR 1.67, 95% CI: 1.56-1.79), and participants in the highest aUFR quartile had a 5.39-fold greater risk than those in the lowest quartile (95% CI: 3.97-7.33). Restricted cubic spline analysis demonstrated a significant nonlinear dose-response relationship, with CKD risk increasing more steeply above an aUFR of 3.80. Longitudinal analyses identified three distinct UFR trajectory patterns; compared with the increasing trajectory, the decreasing trajectory (HR 0.78, 95% CI: 0.63-0.95) and stable low-level trajectory (HR 0.31, 95% CI: 0.25-0.39) were associated with significantly lower CKD risk. The prediction model showed good discrimination (AUC 0.73, 95% CI: 0.71-0.76), and the findings remained robust in temporal validation and sensitivity analyses.

CONCLUSION

The two-year average uric acid-to-eGFR ratio was independently associated with incident CKD and showed good discrimination for risk prediction. The identified nonlinear risk pattern and threshold may provide useful information for risk assessment, although further external validation is required before clinical application.