Accuracy of GFR Change Over Time Using the CKD-EPI Equation

Managing patients with chronic kidney disease requires being able to track progression of disease over time.  The CKD-EPI equation is increasingly used to estimate glomerular filtration rate (GFR).  This equation primarily uses serum creatinine as a kidney function marker, and incorporates age, race, and sex to account for the non-GFR determinants of serum creatinine.  Age, race, and sex all serve as a proxy for muscle mass, the primary source of creatinine generation, and if muscle mass changes over time, a once accurate GFR estimate could become inaccurate.  Accordingly, a study published in the August issue of the American Journal of Kidney Diseases evaluated the accuracy of the CKD-EPI equation over time in a population including people with and without CKD. Corresponding author Dr. Lesley Inker (LI) at Tufts Medical Center discussed her findings with eAJKD contributor and AJKD Deputy Editor, Dr. Daniel Weiner (eAJKD).

eAJKD: Can you briefly summarize what this study showed?

LI: To look at the question of how the CKD-EPI creatinine-based GFR estimating equation performed over time, we pooled data from four clinical trials, all of which had longitudinal measures of both measured GFR and serum creatinine that were performed on the same day.  We then compared the CKD-EPI estimated GFR to the measured GFR at all of these time points, and showed that the accuracy of the GFR estimates did not change over time.

eAJKD: What does this result mean for clinical care?

LI: In a clinic setting, many times we notice that in certain patients, the estimated GFR changes over a short period of time.  We wondered if that is a valid conclusion?  It’s possible that the estimated GFR today is different than it was a month ago or a year ago because of the non-GFR determinants of creatinine.  Perhaps the person ate more meat, or they lost weight, or they had an amputation.

Can you interpret a change in estimated GFR as a change in the true measured GFR?  Our results suggest that for the most part, clinician can feel confident that if there’s a change in the estimated GFR, it’s related to a change in the measured GFR. There are going to be some people where these non-GFR determinants change so dramatically that it leads to changes in an estimated GFR separate from measured GFR, but for the most part a clinician can interpret any change in the estimated GFR or stability of the estimated GFR as a reflection of what was going on with the measured GFR.

eAJKD: Were you able to look at some of the changes in these non-GFR determinants of creatinine over time?

LI: Unfortunately, we were not.  We did not have sufficient data to look at changes in these variables over time.

eAJKD: Were there changes in the association that you saw between estimated GFR and measured GFR based on extremes in body habitus or age, as these are some of the other determinants that we know may affect the relationship between creatinine and GFR?

LI: We looked at the results by age, sex, race, and BMI at baseline, and we were reassured that there did not appear to be any differences.

eAJKD: In the study, you used iothalamate GFR that was conducted as part of these major clinical trials as your gold standard.  Can you explain how that’s performed?

LI: When you measure GFR, the goal is to assess what’s going on in the filtration component of the kidney, and not any of the tubular secretion or reabsorption. Therefore, you want to give an exogenous marker that is completely filtered by the kidney and not reabsorbed or secreted.  Iothalamate is one example of such an exogenous marker.  In these studies, participants were given iothalamate and then the urinary clearance of iothalamate was measured.

For all of these studies, the iothalamate is given subcutaneously and then the investigators waited for about an hour to let it distribute throughout the body.  Then iothalamate was measured twice in the blood, 30 minutes apart.  During those 30 minutes, a urine collection was made.  By comparing the average plasma level from those two time points against how much iothalamate was excreted in the urine, the kidney clearance rate for iothalamate can be calculated.

In each of these studies, they did three or four of these urine collection periods, and then the investigators averaged those together to get the GFR over about a three to four hour time period.

eAJKD: Given the complexity of the gold standard iothalamate measurement, do you think that there’s a role for determining measured GFR in clinical practice?

LI: In clinical practice, the estimated GFR is much easier to do.  There’s much less room for error with estimated GFR so that it could even be more accurate than the measured GFR.

The estimated GFR was developed in specific populations.  For patients who are very different from the populations in which the estimating equation was developed, the estimate could be very different from the true measured GFR.

To use an extreme example, in a patient with an amputation, the muscle mass is going to be very different from a person of the same age, sex, and race who was included in the population in which the equation was developed.  In that example, it would be worthwhile to have a measured GFR because it will be much closer to the true GFR than from any estimate.

eAJKD: Once you have one measured GFR, do you think it’s important to repeat that over time?

LI: Because the measured GFR has some random error, by averaging across multiple measures, even over time, you’re going to get closer to the truth.  This is more complicated when you have to give an exogenous marker and also collect multiple urine specimens. What many of us more commonly do in practice is use a 24-hour urine collection to measure creatinine clearance. This of course is really only an estimate of the measured GFR, but is not biased by the non-GFR determinants like estimates based on serum creatinine will be.  At the same time, no measure is perfect as doing a 24-hour collection is also prone to error.  Just like with iothalamate and estimated GFR using serum creatinine, having more than one 24-hour collections will reduce the error from any one measurement.

(Note: there is a recent article on a related topic that appeared in NEJM.)

To view the article abstract or full-text (subscription required), please visit

Leave a Reply

%d bloggers like this: