An investigation into the accuracy of using serum creatinine estimated glomerular filtration rate to predict measured glomerular filtration rate

Abstract

Objectives: The aim of the study was to identify a threshold value of the estimated glomerular filtration rate (eGFR) that would predict a measured glomerular filtration rate (mGFR) of ≤25 ml/min/1.73 m. This is to guide use of the new British Nuclear Medicine Society guidelines, which specifies a 24 h sample for patients with a GFR of ≤25 ml/min/1.73 m, when using the single-sample (SS) method.

Patients and methods: Data from 1956 patient studies were used to calculate values for the GFR using the slope intercept (SI) method and SS , which was taken 210 min after injection. A sub-data set of 241 patients was taken with patients having an mGFR of ≤80 ml/min/1.73 m by the SS or SI method and an eGFR value within 28 days of the mGFR. The Modification of Diet in Renal Disease equation was used to calculate the eGFR. Receiver operator characteristic curves comparing the sensitivity and specificity of using the eGFR to predict values of the mGFR were used to find a threshold value for the eGFR.

Results: There is a large variation in the accuracy of using the eGFR to predict the mGFR value. The mean difference for the SI method is -2%, with an absolute percentage difference of 22% and a range of -121 to 63%. The SS method has a mean percentage difference of -1%, absolute percentage difference of 23% and a range of -153 to 86%.From the receiver operator characteristic graphs, an eGFR value of 40 ml/min/1.73 m was chosen to maximise the sensitivity of the eGFR threshold value used to predict an mGFR value of ≤25 ml/min/1.73 m.

Conclusion: The eGFR value has a high error when used for predicting the mGFR. Therefore, the study recommends that an eGFR threshold value of ≤40 ml/min/1.73 m is used to predict mGFRs of ≤25 ml/min/1.73 m in order to optimise the likelihood of identifying all such patients.