Václav Ptáčník1, David Zogala1, Daniela Skibová1,2, Hana Jiskrová1, Jiří Trnka1,2, Vladimír Tesař3, Romana Ryšavá3, Martin Šámal4. 1. Institute of Nuclear Medicine, First Faculty of Medicine, Charles University and General University Hospital in Prague, Salmovská 3, 120 00, Praha 2, Czech Republic. 2. Department of Radiation Protection, General University Hospital in Prague, U nemocnice 2, 128 08, Praha 2, Czech Republic. 3. Department of Nephrology, First Faculty of Medicine, Charles University and General University Hospital in Prague, U nemocnice 2, 128 08, Praha 2, Czech Republic. 4. Institute of Nuclear Medicine, First Faculty of Medicine, Charles University and General University Hospital in Prague, Salmovská 3, 120 00, Praha 2, Czech Republic. samal@cesnet.cz.
Abstract
OBJECTIVES: Information on renal function required before specified radiological examinations with contrast agents is usually obtained through prediction equations using serum creatinine and anthropometric data. The aim of our study was to demonstrate discrepancy between poor prediction and good diagnostic accuracy of glomerular filtration rate (GFR) estimated by prediction equations. METHODS: In 50 patients, reference GFR was measured as plasma clearance of 51-chromium labeled ethylene-diamine-tetraacetic-acid (51Cr-EDTA) and compared with GFR assayed by creatinine clearance (CC) and estimated by Cockcroft-Gault prediction equation (CG). For comparisons, CC and CG were considered as continuous, categorical, and binary variables. Accuracy of the reference GFR prediction was expressed in terms of prediction errors and diagnostic accuracy indices. RESULTS: As continuous variable, CG estimated individual values of GFR with large prediction error exceeding that of CC. As categorical variable, it classified the patient stage of chronic kidney disease (CKD) with medium diagnostic accuracy of 74% (CKD 3) and 62% (CKD 4). As binary variable, CG classified individual patient's GFR below 30 and 60 ml/min/1.73 m2 with good diagnostic accuracy of 80 and 94%, respectively. Performance of other prediction equations did not significantly differ from CG. CONCLUSIONS: Despite large variance and poor prediction accuracy of individual GFR estimates, most of them correctly classified individual patient's GFR below specified level. Results of prediction equations thus should be used and reported exclusively as binary variables, while numerical values of GFR, if required, should be measured by more accurate radionuclide or laboratory methods. KEY POINTS: • Radiological guidelines on contrast media require estimation of glomerular filtration rate to assess kidney function before specified contrast examinations. • Estimated glomerular filtration rate is obtained through prediction equations using serum creatinine and anthropometric data as predictors. • While numerical estimates of glomerular filtration rate are inaccurate (their prediction accuracy is poor), diagnostic accuracy of binary estimates (ability to classify patient's glomerular filtration rate below or above a specified level) is very good.
OBJECTIVES: Information on renal function required before specified radiological examinations with contrast agents is usually obtained through prediction equations using serum creatinine and anthropometric data. The aim of our study was to demonstrate discrepancy between poor prediction and good diagnostic accuracy of glomerular filtration rate (GFR) estimated by prediction equations. METHODS: In 50 patients, reference GFR was measured as plasma clearance of 51-chromium labeled ethylene-diamine-tetraacetic-acid (51Cr-EDTA) and compared with GFR assayed by creatinine clearance (CC) and estimated by Cockcroft-Gault prediction equation (CG). For comparisons, CC and CG were considered as continuous, categorical, and binary variables. Accuracy of the reference GFR prediction was expressed in terms of prediction errors and diagnostic accuracy indices. RESULTS: As continuous variable, CG estimated individual values of GFR with large prediction error exceeding that of CC. As categorical variable, it classified the patient stage of chronic kidney disease (CKD) with medium diagnostic accuracy of 74% (CKD 3) and 62% (CKD 4). As binary variable, CG classified individual patient's GFR below 30 and 60 ml/min/1.73 m2 with good diagnostic accuracy of 80 and 94%, respectively. Performance of other prediction equations did not significantly differ from CG. CONCLUSIONS: Despite large variance and poor prediction accuracy of individual GFR estimates, most of them correctly classified individual patient's GFR below specified level. Results of prediction equations thus should be used and reported exclusively as binary variables, while numerical values of GFR, if required, should be measured by more accurate radionuclide or laboratory methods. KEY POINTS: • Radiological guidelines on contrast media require estimation of glomerular filtration rate to assess kidney function before specified contrast examinations. • Estimated glomerular filtration rate is obtained through prediction equations using serum creatinine and anthropometric data as predictors. • While numerical estimates of glomerular filtration rate are inaccurate (their prediction accuracy is poor), diagnostic accuracy of binary estimates (ability to classify patient's glomerular filtration rate below or above a specified level) is very good.
Entities:
Keywords:
Contrast media; Creatinine; Glomerular filtration rate; Kidney function tests
Authors: Andrew D Rule; Timothy S Larson; Erik J Bergstralh; Jeff M Slezak; Steven J Jacobsen; Fernando G Cosio Journal: Ann Intern Med Date: 2004-12-21 Impact factor: 25.391
Authors: Andrew S Levey; Josef Coresh; Tom Greene; Lesley A Stevens; Yaping Lucy Zhang; Stephen Hendriksen; John W Kusek; Frederick Van Lente Journal: Ann Intern Med Date: 2006-08-15 Impact factor: 25.391
Authors: J Björk; S-E Bäck; G Sterner; J Carlson; V Lindstrom; O Bakoush; P Simonsson; A Grubb; U Nyman Journal: Scand J Clin Lab Invest Date: 2007 Impact factor: 1.713
Authors: David Zogala; Václav Ptáčník; Irena Maříková; Daniela Chroustová; Martin Šámal Journal: Eur J Nucl Med Mol Imaging Date: 2022-05-28 Impact factor: 10.057