Propagation of measurement errors in glomerular filtration rate determination: a comparison of slope-intercept,...

Measurement errors occurring during glomerular filtration rate (GFR) studies propagate to an error in the calculated GFR. Previous work has modelled measurement errors for slope-intercept (SI-GFR), single-sample (SS-GFR) and slope-only (SO-GFR) methods. In this study, we have extended these models. The primary aims were to (i) compare measurement errors in two-sample SI-GFR, three-sample SI-GFR, SS-GFR and SO-GFR, and (ii) determine the sensitivity of GFR to errors arising from different measurements.

This study expanded on previous models of GFR measurement error incorporating biological data from 786 patients and realistic measurement errors. GFR median absolute error and the coefficient of variation (CV) were calculated for each method. A sensitivity analysis was carried out for individual measurement errors.

The median absolute error ranged between 1.2 and 2.3 ml/min/1.73 m, lowest for SS-GFR (4 h) and highest for SO-GFR. At higher rates of clearance, CV was less than 5% for all methods. CV increased rapidly when GFR decreased below a threshold ranging between 34 and 56 ml/min/1.73 m, lowest for three-point SI-GFR and highest for SO-GFR. SI-GFR and SS-GFR are most sensitive to injected activity quantification, but less sensitive to other measurement errors.

Measurement errors are probably insignificant relative to biological variation for GFR of more than 60 ml/min/1.73 m, but become significant irrespective of biological variation at lower GFR, particularly in serial studies when GFR less than 25 ml/min/1.73 m. Limits of precision recommended in the 2018 British Nuclear Medicine Society guideline are appropriate for once-off GFR measurement, whereas slightly more stringent limits are proposed for serial studies.