Comprehensive Kidney Function Assessment and Early Warning in Older Adults

Overview

This completed observational cohort study evaluates a comprehensive set of clinical and molecular markers of kidney function in two established cohorts of Chinese adults aged 60 years or older. The study assesses markers reflecting glomerular filtration, kidney tubular transport, and kidney endocrine function and evaluates their associations with kidney failure, cardiovascular and cerebrovascular events, and all-cause mortality. It also compares approaches to estimating glomerular filtration rate and develops and internally validates an early-warning model for kidney function decline. Existing baseline and follow-up data and stored serum and urine specimens are used. No intervention or additional study-specific clinical procedure is administered.

The study includes 5,588 participants, comprising 1,788 participants from the Rugao Longevity and Ageing Study and 3,800 participants from the Shanghai Older-Adult Cohort. The study uses existing demographic, lifestyle, clinical, and laboratory data, longitudinal follow-up information, and measurements obtained from stored serum and urine specimens. No intervention is assigned, and no additional study-specific examination or biospecimen collection is performed. Markers reflecting glomerular filtration, kidney tubular transport, and kidney endocrine function are evaluated in relation to incident kidney failure, cardiovascular and cerebrovascular events, and all-cause mortality. Approaches to estimating glomerular filtration rate and combinations of kidney function-related biomarkers are compared for prognostic assessment. Multivariable linear, logistic, and Cox proportional hazards regression models are used, as appropriate, to evaluate associations while accounting for potential confounding factors. For prediction modeling, the analytic sample is divided in a 7:3 ratio into model-development and internal-validation sets. Statistical learning and machine-learning methods, including random forests, decision trees, support vector machines, and neural-network approaches, are evaluated. Cross-validation is used during model development and model selection. The study also supports the development of a secure, access-controlled data-sharing and analytics platform.

Conditions