Abstract P018: Calibration of Analytes Over Twenty-Five Years in the Atherosclerosis Risk in Communities Study
Background: Comparability of laboratory measures over time is important for studies of disease prevalence and progression. While a small amount of bias may seem negligible on an individual level, it can result in substantial misclassification of disease in the population. We conducted a calibration study of important biomarkers across five study visits (25 years) in ARIC.
Methods: We re-measured 15 analytes in 200 blood samples to calibrate original measurements at each time point using Bland-Altman plots and Deming regression. We also assessed the impact of calibration on the prevalence of chronic kidney disease (CKD), defined by estimated glomerular filtration rate using creatinine (eGFRcr), and on trends over time.
Results: Assays in samples frozen 12-27 years were highly correlated with original values (median r=0.95) after removing outliers (median 4% of values). The range of bias (% difference in means) across visits for each original analyte compared to its reference were: creatinine: 13-49%; uric acid: 3-24%; C-reactive protein: 3-9%; total cholesterol: 1-6%; high density lipoprotein cholesterol: 4-8% (but new methods differed); low density lipoprotein cholesterol: 1-5%; triglycerides: 2-4%; glucose: 1-4%; N-terminal prohormone of brain natriuretic peptide: 2-12%; high sensitivity cardiac troponin T: 1-9%; alanine transaminase (ALT): 21%; aspartate transaminase (AST): 17%; gamma glutamyl transpeptidase: 0.2%; ß2-microglobulin: 1%; beta-trace protein: 13%. Four analytes met calibration criteria: creatinine, uric acid, ALT and AST. The impact on CKD prevalence was substantial and similar to previous statistical calibration (22% uncalibrated, 1.9% previously and 1.3% current laboratory calibration). Trends in eGFRcr over time were better aligned after calibration (Figure).
Conclusions: Repeat assay of samples shows high correlation with original values. Calibration enables application of absolute cutoffs (required for defining CKD and other conditions) and improves longitudinal analyses.
Author Disclosures: C.M. Parrinello: None. M.E. Grams: None. D. Couper: None. C.M. Ballantyne: B. Research Grant; Modest; grant support from Roche Diagnostics. H. Other; Modest; investigator on a provisional patent filed by Roche for use of biomarkers in heart failure prediction. R.C. Hoogeveen: B. Research Grant; Modest; grant support from Roche Diagnostics. H. Other; Modest; investigator on a provisional patent filed by Roche for use of biomarkers in heart failure prediction. J.H. Eckfeldt: None. E. Selvin: None. J. Coresh: None.
- © 2014 by American Heart Association, Inc.