Abstract
INTRODUCTION: KIM-1, TNFRSF1A, and TNFRSF1B have been accepted as early risk markers in diabetic kidney disease by the US Food and Drug Administration. Whether they may be useful in identifying high-risk patients for cardiovascular/kidney clinical trial enrollment in other important subgroups is uncertain. METHODS: We evaluated the potential prognostic enrichment of KIM-1, TNFRSF1A, and TNFRSF1B in four cohorts: the Atherosclerosis Risk in Communities (ARIC) (N = 4,594, mean age 76 years, 55% women, mean eGFR 68 mL/min/1.73 m2), African American Study of Kidney Disease and Hypertension (AASK) (N = 705, mean age 55 years, 39% women, mean mGFR 46 mL/min/1.73 m2), Chronic Renal Insufficiency Cohort (CRIC) (N = 2,943, mean age 59 years, 45% women, mean eGFR 35 mL/min/1.73 m2), and Boston Kidney Biopsy Cohort (BKBC) (N = 434, mean age 54 years, 48% women, mean eGFR 51 mL/min/1.73 m2). We evaluated three outcomes: 40% glomerular filtration rate (GFR) decline, kidney failure, and incident cardiovascular disease (CVD) overall and in two subgroups historically underrepresented in clinical trials: participants with no diabetes, and those with albuminuria <200 mg/g. RESULTS: Published models (40% decline tool, kidney failure risk equation, and PREVENT) using clinical variables had moderate to strong risk discrimination in each cohort: 40% GFR decline, AUROC range: 0.78-0.90; kidney failure, C-statistic range: 0.75-0.93; and CVD, C-statistic range: 0.59-0.79. After addition of biomarkers, there was a small but significant improvement in the meta-analyzed overall population: change in AUROC in 40% GFR decline: 0.02, p < 0.001; change in C-statistic for kidney failure: 0.01, p = 0.02; change in C-statistic for CVD: 0.01, p = 0.03. Among participants without diabetes, the change was statistically significant only for 40% decline; among patient with albuminuria <200 mg/g, the change was statistically significant only for the two kidney outcomes. CONCLUSION: KIM-1, TNFRSF1A, and TNFRSF1B may not be strong prognostic enrichment biomarkers over and above clinical risk estimates. Clinical trials should test whether they help with predictive enrichment.