What eGFR estimates and why creatinine is used to do it
Glomerular filtration rate is the volume of plasma the glomeruli filter each minute, and it is the single best summary of kidney function. Measuring it directly means infusing a marker cleared purely by filtration — inulin, iohexol, iothalamate — and sampling repeatedly. That is a research procedure, not a clinic test, so filtration is estimated instead from an endogenous marker whose production is reasonably steady.
Creatinine is that marker. Muscle produces it at a rate that is fairly constant for a given person, the glomerulus filters it freely, and it is cheap to measure. What creatinine cannot do is tell an equation how much muscle a person has. Two people with the same filtration rate but different muscle mass will have different creatinine concentrations, so every creatinine-based equation must guess at muscle mass from demographic proxies. CKD-EPI 2021 uses age and sex for that job, and nothing else.
The result is reported per 1.73 m² of body surface area — the average adult surface area used when the convention was set in the 1920s. Indexing lets you compare one patient with another and with a population reference. It also means the reported figure is not that patient's actual filtration rate in mL/min unless their surface area happens to be 1.73 m², which is why this calculator also reports the de-indexed value.
How the CKD-EPI 2021 equation is built
The equation has four multiplicative parts, and each one does a specific job.
The two-limb creatinine term. Rather than one power of creatinine, CKD-EPI uses a breakpoint κ — 0.7 mg/dL for women, 0.9 mg/dL for men — and applies a gentle exponent below it and a steep one (−1.200) above it. Written as min(Scr/κ, 1)α × max(Scr/κ, 1)−1.200, only one limb is ever active: below the breakpoint the second factor is 1, above it the first factor is 1. This spline is what fixed MDRD's well-known problem of underestimating GFR in people with near-normal creatinine.
The age term. 0.9938 raised to the age, a decline of about 0.62% per year, reflecting the fall in filtration and muscle mass with age.
The sex terms. Sex enters three times: through κ, through α (−0.241 for women against −0.302 for men), and through a final multiplier of 1.012 for women. These absorb the systematically lower creatinine production of women at any given filtration rate.
What is not in it. There is no race coefficient. The 2009 CKD-EPI equation multiplied results for Black patients by 1.159, on the empirical observation that creatinine ran higher in that group. A joint National Kidney Foundation and American Society of Nephrology task force concluded in 2021 that race is a social construct and should not be used in clinical algorithms, and the refit 2021 equation dropped it. The practical consequence is that eGFR values for Black patients are modestly lower under the 2021 equation than under the 2009 one — an intended change that moves some patients into an earlier CKD category and earlier onto transplant waiting lists.
The combined creatinine–cystatin C equation adds a second, independent marker. Cystatin C is produced by all nucleated cells and is far less dependent on muscle, so combining the two markers cancels much of the error each carries alone. It is the most accurate of the routinely available estimates and is what KDIGO recommends when a creatinine-based result will change a decision.
Worked example: a 65-year-old man with a creatinine of 1.2 mg/dL
Take the calculator's default patient and follow the arithmetic.
- Form the ratio. κ is 0.9 for a man, so Scr/κ = 1.2 ÷ 0.9 = 1.3333.
- Pick the limb. The ratio is above 1, so min(ratio, 1) = 1 and that term contributes nothing. Only the −1.200 limb is active.
- Evaluate it. 1.3333−1.200 = 0.70806.
- Age term. 0.993865 = 0.66747.
- Multiply. 142 × 0.70806 × 0.66747 = 67.1 mL/min/1.73 m². No female multiplier applies.
- Assign the category. 67 sits in the 60 to 89 band, so KDIGO category G2.
- De-index it. With a height of 175 cm and a weight of 80 kg, Mosteller body surface area is √(175 × 80 ÷ 3600) = √3.889 = 1.972 m². The absolute clearance is 67.1 × 1.972 ÷ 1.73 = 76.5 mL/min.
Two lessons sit in that last step. First, this larger-than-average man's kidneys filter about 76 mL of plasma a minute, not 67 — the indexed number is smaller purely because he is bigger than the 1.73 m² reference. Second, the indexed and de-indexed figures answer different questions, and mixing them up in a dosing decision is a real source of error. For dosing, most labels want a Cockcroft-Gault clearance in mL/min, which is a different equation again.
Reading the number: category, chronicity and albuminuria
KDIGO defines six GFR categories, from G1 at 90 or above to G5 below 15. The categories are thresholds on a continuum, not diagnoses, and two things must be true before an eGFR becomes a diagnosis of chronic kidney disease. The abnormality must have persisted for more than three months — a single low value can reflect dehydration, a drug, or an acute injury. And staging is two-dimensional: the GFR category is paired with an albuminuria category from a urine albumin-to-creatinine ratio, A1 below 30 mg/g, A2 from 30 to 300, A3 above 300. A patient at G2 with A3 albuminuria is at substantially higher risk than one at G3a with A1, which is precisely why the guideline refuses to reduce kidney disease to a single number.
An eGFR of 60 or above with no albuminuria, no structural abnormality and no urinary sediment is not chronic kidney disease. An eGFR that has fallen by 25% or more from a documented baseline is significant even if the absolute value is still above 60.
Precision matters when you read a single result. In the CKD-EPI 2021 development cohorts, roughly 85 to 90% of creatinine-based estimates fell within 30% of measured GFR. That means an estimate of 60 is compatible with a true GFR somewhere near 45 to 78 in the great majority of cases. Treat eGFR as a band, not a point, and watch the trajectory across several results rather than reacting to one.
The situations in which creatinine-based estimation misleads are worth memorising: extremes of muscle mass in either direction, amputation and paralysis, cirrhosis, rapidly changing renal function, high-protein or creatine supplementation, and drugs such as trimethoprim and cimetidine that block tubular creatinine secretion without touching filtration. In each, cystatin C or a combined estimate is the better answer.
eGFR by age and creatinine, CKD-EPI 2021
| Serum creatinine | Age 40 | Age 55 | Age 70 | Age 85 |
|---|---|---|---|---|
| 1.0 mg/dL | 102.8 | 93.7 | 85.4 | 77.8 |
| 1.2 mg/dL | 82.9 | 75.6 | 68.9 | 62.8 |
| 1.5 mg/dL | 63.4 | 57.8 | 52.7 | 48.0 |
| 2.0 mg/dL | 44.7 | 40.7 | 37.1 | 33.8 |
| 3.0 mg/dL | 27.5 | 25.1 | 22.9 | 20.8 |
| 5.0 mg/dL | 14.9 | 13.6 | 12.4 | 11.3 |
Multiply any cell by 1.012 and recompute with κ = 0.7 and α = −0.241 for women; the female values are not a fixed percentage of the male ones because the breakpoint differs.
Which equation for which job
CKD-EPI 2021 creatinine for routine reporting and KDIGO staging. CKD-EPI 2021 creatinine–cystatin C when the creatinine-based value is doubtful or when confirming a result that will change management — assessing transplant eligibility, starting a drug with a hard renal cut-off, or explaining a mismatch between the number and the patient. Cockcroft-Gault in mL/min for renal drug dosing, because product labelling was written against it. Measured clearance — iohexol, or a timed urine collection — when the estimate cannot be trusted at all, as in an amputee, a body-builder, or a patient with extremely low muscle mass.
Mistakes that produce a misleading eGFR
- Using it during acute kidney injury. Creatinine lags filtration by a day or more, so an estimate taken while function is falling reports a GFR the patient no longer has.
- Trusting a single value. Chronic kidney disease requires abnormality persisting beyond three months. One result is a screening finding.
- Reporting GFR without albuminuria. KDIGO staging needs both axes; the urine albumin-to-creatinine ratio carries independent prognostic weight.
- Applying it to a patient with atypical muscle mass. Amputation, paralysis, cirrhosis and body-building all break the assumption that creatinine tracks filtration.
- Dosing drugs from an indexed eGFR. mL/min/1.73 m² is a comparison unit. De-index it, or better, calculate Cockcroft-Gault.
- Comparing an old MDRD result with a new CKD-EPI one. The two disagree systematically above 60, so an apparent improvement may be an equation change.
- Using an adult equation in children. Under 18, use the CKiD paediatric equations, which include height.
- Ignoring the assay. The equation assumes IDMS-standardised creatinine. A non-standardised assay reads high and pushes eGFR down.
What to calculate alongside it
A reduced eGFR rarely travels alone. Advanced chronic kidney disease disturbs mineral metabolism, so albumin-corrected calcium belongs in the same review, and uraemic acidosis shows up as a raised anion gap. When creatinine has risen acutely rather than chronically, the fractional excretion of sodium helps separate a prerenal picture from established tubular injury, and a raised osmolar gap alongside a raised anion gap points at a toxic ingestion rather than kidney disease itself.
On the therapeutic side, once you have a clearance you still need a dose. The weight-based dosage calculator and the body surface area calculator cover the arithmetic that follows, and the latter uses the same Mosteller formula this page uses to de-index eGFR.