Poster #62: Luke Ni

The Clinical and Genetic Landscape of CT-Derived Kidney Biomarkers

Luke Ni, Cameron Beeche, Rakesh Sharma, David Zhang, PhD, Jeffrey T. Duda, PhD, Arijitt Borthakur, PhD, MBA, James C. Gee, PhD, Dan Rader, MD, Charles E. Kahn, Jr., MD, MS, Walter R. Witschey, PhD, Hersh Sagreiya, MD

Chronic kidney disease (CKD) is a progressive kidney disease diagnosed utilizing estimated glomerular filtration rate (eGFR) and albuminuria, which are subject to biases. However, routine abdominal computed tomography (CT) scans contain rich, underutilized anatomical data. This study examines kidney image-derived phenotypes (IDPs) as renal biomarkers in a large, ancestrally diverse cohort, exploring their links to demographic, genetic, and clinical associations. Methods: Clinical, genetic, and phenotypic traits were extracted from an academic biobank cohort of 19,790 patients. Total kidney volume (TKV) and renal cyst volume (RCV) were first segmented via the deep learning TotalSegmentator tool. Subsequent regression analyses were performed across urine and blood biomarkers, including eGFR, a1c, hemoglobin, creatinine, and blood urea nitrogen. Phenome-wide and genome-wide association studies (PheWAS, GWAS) were conducted using TKV and RCV as predictors (R v4.3, Plink 2.0). Analyses controlled for age, sex, and five principal components. Height and weight were used as covariates in analyses excluding the RCV GWAS. Results: Regression models demonstrated significant associations between TKV/RCV and core biomarkers (p < 5e-5). In PheWAS analyses of 2,716 traits (337 and 317 Bonferroni-significant traits in TKV and RCV, respectively; p < 1.84e-5), elevated TKV was associated with liver, bladder, and pancreas neoplasms and gastrointestinal pathologies (fatty liver disease, hernias, pancreatitis), and elevated RCV was correlated with most significant traits. Lowered TKV was linked with cardiorenal diseases such as CKD, renal failure, hypertension, and heart disease. In the GWAS, TKV was linked to genes RARB and FIBCD1 and SNPs near SOX2 and STC1. RCV was significantly associated with BCAS3. Conclusions: This study shows how TKV and RCV from routine clinical imaging are robustly linked to, genetic loci, disease phenotypes, and clinical indicators in a large, ancestrally diverse biobank. Leveraging these broad and comprehensive clinical IDPs can provide clinical and genetic insights that advance population health.