Search

Purpose: In theory, radionuclide measurement of glomerular filtration rate (GFR) using bolus injection and a single blood sample results in GFR values close to the results of slope–intercept (SI) method using two or more blood samples, especially if the sample is collected at specific time. The aim of our study was to provide experimental validation of this assumption, to design the new single-sample methods potentially correcting for inadequate sampling time, and to validate them together with the method recommended by the British Nuclear Medicine Society (BNMS) guidelines. Methods: New analytic and regression-based single-sample methods were developed using previously published concepts. In 373 patients, GFR measured by the new methods and by the method recommended by the BNMS guidelines was compared with the reference GFR measured by SI method using two blood samples. Prediction of the reference GFR was evaluated by cross-validation with all blood samples and with blood samples collected at time interval ± 30 min around optimal sampling time. Results: With all 768 blood samples (regardless of sampling time), the method recommended by the BNMS guidelines performed significantly better than the new methods, predicting reference GFR with the mean absolute error (MAE) of 2.55 (versus 2.95–3.37) ml/min/1.73 m2. In the subset of blood samples collected at time interval ± 30 min around optimal sampling time derived from SI method (n= 247), performance of all methods was significantly improved. MAE was reduced to 1.18 (vs. 1.27–1.39) ml/min/1.73 m2. Realistic estimate of optimal sampling time using creatinine prediction equation CKD-EPI (n= 235) resulted in MAE of 1.37 (vs. 1.53–1.76) ml/min/1.73 m2. Conclusion: The study experimentally confirmed that collection of single blood sample near patient-specific optimal sampling time significantly improves accuracy of prediction of the reference SI GFR by all validated single-sample methods. The method recommended by the BNMS guidelines provided the best results with low prediction error that was further reduced by collecting blood sample at optimal sampling time. The new methods successfully corrected for small deviations from optimal sampling time.