Comparative performance of age-continuous and conventional reference intervals in long-term validation: A 10-year health examination cohort study of serum creatinine.

BACKGROUND: Reference intervals (RIs) are essential for interpreting clinical laboratory results. Conventional RIs are typically defined as fixed adult intervals or discrete age-group-specific intervals, which fail to capture the continuous nature of physiological aging. For biomarkers such as serum creatinine (Cr), this limitation may lead to misclassification, particularly in older adults. Age-continuous reference intervals (ACRIs) have been proposed as an alternative, yet their real-world stability and applicability remain insufficiently validated. METHODS: Using large-scale health examination data from Peking Union Medical College Hospital collected between 2014 and 2023, we established sex-specific ACRIs for serum creatinine. A presumptively healthy reference population was identified using a combined strategy incorporating SOM-based assessment of renal biomarker homogeneity and electronic health record-based clinical eligibility screening. ACRIs were constructed using generalized additive models for location, scale, and shape (GAMLSS) with bootstrap resampling, while conventional sex-specific fixed RIs were derived from the same training dataset using the refineR method. Independent annual health examination cohorts from 2019 to 2023 were used for temporal validation in the overall population and older adults (≥60 years). A sensitivity analysis was additionally performed using expanded blood pressure eligibility criteria. RESULTS: The ACRIs exhibited smooth age-dependent trajectories with narrow confidence intervals and met the predefined validation criterion across all five validation years. In the overall population, ACRIs and conventional RIs showed relatively similar performance. In older adults, however, ACRIs achieved consistently higher within-RI proportions than conventional RIs (96.14% vs 91.83% in females and 95.74% vs 90.37% in males), while conventional RIs failed validation in several year- and sex-specific subgroups. CONCLUSIONS: Age-continuous, sex-specific ACRIs derived from real-world data provide superior temporal stability and applicability compared with conventional fixed RIs for age-related biomarkers such as serum creatinine. This study offers a practical methodological framework for constructing and validating ACRIs using large-scale health examination data.