Accuracy of a Glomerular Filtration Rate Estimating Equation over Time in People with a Wide Range of Kidney Function.
Introduction: The change in glomerular filtration rate (GFR) is important for clinical
decision making. However, the accuracy of GFR estimated from serum creatinine over time
is not well known. The difference between measured GFR (mGFR) and estimated GFR (eGFR)
(error) is usually attributed to non-GFR determinants of serum creatinine. We
hypothesized that the mean error in a populat... read moreion would remain stable over time, but the
inter-individual variation in the change over time in error would be large and related
to clinical and demographic factors associated with non-GFR determinants of serum
creatinine. Methods: This is a longitudinal study of diagnostic accuracy including
subjects from four studies with a wide range of kidney function. GFR was measured using
urinary clearance of 125I-iothalamate (reference test). GFR was estimated using the
Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (index test). The
change in error over time was modeled using longitudinal mixed models. Baseline
covariates hypothesized to be associated with the non-GFR determinants of serum
creatinine were tested in the mixed model. Results: There were 13,708 GFR measurements
in 3635 subjects over a mean follow up period of 3.6 years. In the pooled dataset the
mean measured and estimated GFR and error at baseline were 76, 76, and -0.3 ml/min/1.73
m2. The mean change (standard error) in measured and estimated GFR and error were -2.3
(0.12), -2.2 (0.09) and -0.1 (0.10) ml/min/1.73 m2 per year (P <.0001, <.0001, and
0.6 respectively). The variability (SD) among subjects in changes in measured and
estimated GFR and error was 2.24, 1.59, and 1.91 ml/min/1.73 m2 per year, respectively.
Only 16% of subjects had changes in error larger than + 3 ml/min/1.73 m2 per year. A
total of 8 non-GFR determinants were significantly associated with inter-individual
variation in change in error in at least one study. Of the 8, only 1 explained greater
than 20% of the variation [urine protein (22%)]. Conclusion: The accuracy of GFR
estimates did not change over time in the population. Clinicians should interpret
changes in estimated GFR over time as reflecting changes in measured GFR in most
Thesis (M.S.)--Tufts University, 2011.
Submitted to the Dept. of Clinical & Translational Science.
Advisors: Andrew Levey, and Lesley Stevens.
Committee: Christopher Schmid.
Keyword: Medicine.read less