Guide
eGFR after a creatinine method change: how a small bias moves CKD categories
No patient’s kidneys change on the day the laboratory changes its creatinine method, but every eGFR it reports may. This guide works through the equation to show how a creatinine bias becomes an eGFR shift, why the categories that matter sit where the equation is steepest, and what to check at the switch.
The question a method change raises
eGFR is not measured. It is calculated from serum creatinine with age and sex, and then placed in a GFR category that drives referral, drug dosing and the label of chronic kidney disease. In the categories used in guidelines, G2 runs down to 60, G3a to 45, G3b to 30 and G4 to 15 mL/min/1.73 m², with G1 at 90 or over [NICE NG203]. Any systematic difference between the old and new creatinine methods passes through the equation into every eGFR, and into the category of every patient close to a boundary.
The guide to Jaffe and enzymatic creatinine covers why two methods can differ. This one takes the difference as given and follows it through the equation.
Inside the equation
The CKD-EPI 2021 creatinine equation, without a race term, is [Inker et al., 2021]:
eGFR = 142 × min(Scr/κ, 1)^α × max(Scr/κ, 1)^−1.200 × 0.9938^age × 1.012 if female
Scr is creatinine in mg/dL; κ is 0.7 for women and 0.9 for men; α is −0.241 for women and −0.302 for men [Inker et al., 2021]. The equation has two slopes joined at a knee where creatinine equals κ. Below the knee, eGFR responds weakly to creatinine (exponent α). Above it, eGFR responds strongly (exponent −1.200).
From creatinine bias to eGFR change
Because the equation is a product of powers, a creatinine bias of a given percentage changes eGFR by the same percentage for every patient on the same side of the knee, whatever their age or sex:
eGFR change = (1 + creatinine bias)^exponent − 1
Above the knee the exponent is −1.200, so eGFR moves about 1.2 times as far as creatinine, in the opposite direction. The EKFC equation behaves the same way, with an exponent of −1.132 above its own reference point [Zhao et al., 2023].
| CKD-EPI 2021 | EKFC | |||
|---|---|---|---|---|
| Creatinine bias | Above the knee | Below, men | Below, women | Above Q |
| +2.5% | −2.9% | −0.7% | −0.6% | −2.8% |
| −2.5% | +3.1% | +0.8% | +0.6% | +2.9% |
| +5.0% | −5.7% | −1.5% | −1.2% | −5.4% |
| −5.0% | +6.3% | +1.6% | +1.2% | +6.0% |
| +10.0% | −10.8% | −2.8% | −2.3% | −10.2% |
| −10.0% | +13.5% | +3.2% | +2.6% | +12.7% |
Where the knee sits
At the knee, eGFR depends only on age and sex. Computed from the equation, it is 111 for a man and 112 for a woman at 40; 98 for a man and 99 for a woman at 60; and 86 for a man and 87 for a woman at 80 mL/min/1.73 m². So for adults up to 80, every eGFR below about 86 comes from a creatinine above the knee, on the steep part of the curve. That covers the 60, 45, 30 and 15 boundaries entirely. The weak response below the knee only protects results that are already in G1 or high in G2.
In practice, then, a creatinine method change moves every eGFR near a decision boundary by roughly 1.2 times the creatinine bias at that concentration. A constant creatinine offset matters more than this rule suggests at low creatinine, where it is a larger percentage; a proportional bias moves every result by the same percentage.
From a percentage to a category
Turn the boundaries back into creatinine to see where the method comparison has to be precise. For a 65-year-old, eGFR 60 corresponds to a creatinine of about 1.32 mg/dL (116 µmol/L) in a man and 1.03 mg/dL (91 µmol/L) in a woman. Those are the concentrations at which the creatinine bias needs to be read off the comparison line, not the average bias over the whole range.
Then count. If the new method reads 5% higher, eGFR falls by about 5.7%, and every patient whose eGFR was between 60 and about 63.6 on the old method is now below 60. NICE asks laboratories to report eGFR as a whole number up to 90 [NICE NG203], so rounding moves the exact edge by up to half a unit. How many patients that is depends entirely on how many sit in that narrow band, which is a property of your own population; the equation cannot tell you.
Small for one patient, large for a population
A systematic review of eGFR biological variation puts a minimum conservative reference change value for eGFR at ±12.5%, and an overall value of ±16.5% [Thöni et al., 2022]. Against that, a −5.7% shift from a 5% creatinine bias is less than half the change any single patient would need to show before it counts as real. No individual result looks alarming.
But the shift is not random. It moves every patient in the same direction on the same day, so the whole distribution slides across each boundary and the number of people in each category changes. For a patient being followed over time it also matters: one guideline definition of accelerated progression is a sustained fall in GFR of 25% or more together with a change of category within 12 months [NICE NG203]. A method change can supply the change of category on its own. The current international guideline asks clinicians to assess the potential for error in eGFR when they assess change over time [KDIGO, 2024]; a dated note of the method change is what lets them do it.
One patient, two methods
After calibration was standardised to the reference method, one external quality assessment programme found mean biases of +3.8% for one creatinine peer group and −4.3% for another [Lee et al., 2017]. As an illustration only, take a man of 65 whose creatinine is 1.25 mg/dL on a method that behaves like the second group, and move him to one that behaves like the first. His creatinine reads 8.5% higher; his eGFR goes from 64 to 58 (−9.3%), and his category from G2 to G3a. That is 74% of the eGFR reference change value: inside it, and still a new category on his report.
The eGFR calculator runs this for any patient, with either equation, using these published method differences or the slope and intercept from your own comparison.
What to do at the switch
- Read creatinine bias at the boundary concentrations. Use the comparison line at the creatinine values that correspond to the eGFR boundaries for your typical ages, as above. The guide to method comparison regression covers which line to fit.
- Count the patients who change category. Put each pair’s old and new creatinine through the equation and count the results that land in a different category, set against the number repeat testing alone would move. The change check does this count for the creatinine lines you set, in your browser.
- Keep the equation fixed. Use the same equation across a region [KDIGO, 2024]. If an equation change is also planned, separate it from the method change so each can be measured on its own.
- Prefer a specific creatinine method. Guidelines recommend enzymatic, specific assays traceable to the reference method [KDIGO, 2024] [NICE NG203]. A change towards one is often the reason for the switch, and the reason the bias is not zero.
- Tell the people who follow trends. Put the date and the size of the shift at the boundaries on reports, and give the renal and primary care teams the expected movement before go-live, not after the first progression alert.
Sources
- Inker LA, Eneanya ND, Coresh J, et al. (2021). New creatinine- and cystatin C-based equations to estimate GFR without race. New England Journal of Medicine. pmc.ncbi.nlm.nih.gov/articles/PMC8822996/
- Zhao L, Li H, Liu X, et al. (2023). Validation of the EKFC equation for glomerular filtration rate estimation and comparison with the Asian-modified CKD-EPI equation in Chinese chronic kidney disease patients in an external study. Renal Failure. pmc.ncbi.nlm.nih.gov/articles/PMC9848359/
- National Institute for Health and Care Excellence (2021). Chronic kidney disease: assessment and management (NG203), recommendations. NICE guideline. www.nice.org.uk/guidance/ng203/chapter/Recommendations
- Levin A, Ahmed SB, Carrero JJ, et al. (2024). Executive summary of the KDIGO 2024 Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease: known knowns and known unknowns. Kidney International. kdigo.org/wp-content/uploads/2017/02/KDIGO-2024-CKD-Guideline-Executive-Summary.pdf
- Thöni S, Keller F, Denicolò S, et al. (2022). Biological variation and reference change value of the estimated glomerular filtration rate in humans: a systematic review and meta-analysis. Frontiers in Medicine. pmc.ncbi.nlm.nih.gov/articles/PMC9583397/
- Lee E, Collier CP, White CA (2017). Creatinine assay attainment of analytical performance goals following implementation of IDMS standardization. Canadian Journal of Kidney Health and Disease. pmc.ncbi.nlm.nih.gov/articles/PMC5347424/