Measurement uncertainty in three sample glomerular filtration rate tests

To provide an estimate of the uncertainties associated with the measurement of the glomerular filtration rate using the three-sample method.

Retrospective data from 2063 patients whose glomerular filtration rate (GFR) was evaluated with Tc-99m diethylenetriaminepentaacetic acid using a three-sample method were compiled for this analysis. Input quantities to the GFR measurement were modelled as probability density functions (PDFs), and a Monte-Carlo method of propagating these PDFs was used to assess overall uncertainty in the measured GFR for each patient.

The average uncertainty in the three-sample slope-intercept GFR was found to be 1.53% (with a coverage factor of k  = 1). The dominant source of uncertainty in the slope-intercept GFR was the injected activity measurement using the radionuclide calibrator, except for slope-intercept GFRs below 20 ml/min whose dominant source of uncertainty was counting statistics. The Brøchner-Mortensen correction for the unmeasured initial plasma clearance phase became the dominant source of uncertainty in absolute GFR values above 20 ml/min, increasing the average uncertainty to 3.32% ( k  = 1). The uncertainty in the absolute GFR was found to increase with increasing absolute GFR, because the Brøchner-Mortensen correction was associated with higher uncertainties for high GFRs.

We have developed a framework to assess the measurement uncertainty in three-sample GFR studies for individual patients. While the uncertainty in the slope-intercept GFR is low, empirical corrections applied to derive the absolute GFR increase the uncertainty. Inclusion of measurement uncertainties in clinical GFR reports may aid clinical decision making, particularly in borderline cases.