Research Article: Albuminuria, kidney function, and automatically quantified retinal vascular phenotypes in hospitalized adults with type 2 diabetes
Abstract:
To compare albuminuria and creatinine-based estimated glomerular filtration rate (eGFR) in relation to automatically quantified retinal vascular phenotypes from color fundus photographs in hospitalized adults with type 2 diabetes.
This single-center retrospective study applied clinical, ocular, image-quality, and temporal eligibility criteria. Kidney measurements with a verifiable laboratory-report-to-image interval of no more than 7 days were retained. Exact urinary albumin-to-creatinine ratio (UACR) values were analyzed after log2 transformation; eGFR was recalculated using the 2021 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine equation. Eight vascular outcomes were analyzed at the eye level using patient-clustered robust standard errors. The prespecified primary analyses were separate core models for UACR and eGFR adjusted for age, gender, body mass index, diabetes duration, HbA1c, and systolic blood pressure, with Benjamini–Hochberg false discovery rate (FDR) correction across the eight outcomes within each exposure family. Joint, complete-case, alternative-equation, restricted-range, categorical, spline, and tighter-temporal-window analyses were treated as supportive or exploratory.
The final cohort included 658 eyes from 337 patients; the median absolute laboratory-report-to-image interval was 1 day (IQR, 0–1), and 97.9% of eye records were within 3 days. In the primary separate core models, neither UACR nor 2021 CKD-EPI eGFR was associated with any of the eight vascular outcomes after FDR correction (minimum q = 0.090). UACR showed a nominal association with arterial density ( q = 0.097), and eGFR showed nominal associations with overall, arterial, and 3-mm foveal vessel density. Exploratory complete-case analyses yielded stronger eGFR signals: in the 431-eye/220-patient joint complete-case cohort, the eGFR-only model in the same cohort was FDR-significant for all eight outcomes, and seven remained significant after mutual adjustment with UACR. Alternative-equation and restricted-range analyses also yielded eGFR-related associations, whereas same-day analyses were imprecise and null. In a post hoc smoking-adjusted joint model, however, none of the eight eGFR associations remained significant after FDR correction.
The primary analyses did not demonstrate FDR-significant associations of either kidney marker with the retinal vascular phenotypes. Exploratory complete-case and sensitivity analyses suggested a model-dependent eGFR-related pattern, whereas UACR remained largely null. These cross-sectional findings are hypothesis-generating and require independent longitudinal validation.
Introduction:
Diabetic kidney disease and diabetic retinopathy are common microvascular complications of type 2 diabetes and remain major causes of kidney failure and visual impairment, respectively ( 1 – 8 ). Kidney involvement is assessed primarily using the urinary albumin-to-creatinine ratio (UACR) and estimated glomerular filtration rate (eGFR). These measures provide complementary rather than interchangeable information. UACR reflects increased glomerular permeability and urinary albumin excretion, whereas eGFR estimates…
Read more