MDRD vs. CKD-EPI vs. Mayo Quadratic: Why GFR Formulas Disagree

Three peer-reviewed, published formulas exist for estimating kidney function from a single creatinine blood test, and they routinely disagree with each other for the exact same patient, sometimes by a wide margin. This is not a bug in any calculator or a sign that one of the formulas is simply wrong; it is a well documented, verifiable property of how MDRD, CKD-EPI and the Mayo Quadratic formula were each built. This article runs all three across four different patients, with every number independently hand-verified, to show exactly how large that disagreement gets, and, more usefully, exactly when it shrinks to almost nothing.

MDRD, CKD-EPI and Mayo Quadratic: what each one actually calculates

MDRD, short for Modification of Diet in Renal Disease, the study it was derived from, is the oldest of the three still in common reference use. Its IDMS-traceable four-variable form is:

eGFR = 175 × SCr^-1.154 × Age^-0.203 × 0.742 (if female) × 1.212 (if Black)

CKD-EPI, from the Chronic Kidney Disease Epidemiology Collaboration, was developed afterward specifically to improve on MDRD's known weaknesses. Its original 2009 version included a race coefficient similar in spirit to MDRD's; a 2021 update, now the version most current clinical guidelines and this site's own calculator default to, removed that coefficient entirely after a joint task force of major U.S. kidney and laboratory medicine organizations found it was not well justified and could delay care for Black patients. The 2021 version:

eGFR = 142 × min(SCr ÷ kappa, 1)^alpha × max(SCr ÷ kappa, 1)^-1.2 × 0.9938^Age × sex factor

with kappa = 0.9, alpha = -0.302, sex factor = 1 for men, and kappa = 0.7, alpha = -0.241, sex factor = 1.012 for women.

Mayo Quadratic, published by Rule and colleagues in 2004, was built for a specific, narrower purpose: correcting MDRD's tendency to underestimate GFR in patients with normal or near-normal kidney function. It uses no race term at all, and floors creatinine at 0.8 mg/dL for any value below that:

eGFR = exp(1.911 + 5.249 ÷ SCr − 2.114 ÷ SCr² − 0.00686 × Age − 0.205 (if female))

All three take the same three or four raw inputs and produce a number in the same unit, mL/min/1.73m². That is precisely what makes the disagreement between them worth taking seriously: it is not a difference of units or definitions, it is a genuine difference in what number each formula's underlying statistical fit produces from identical inputs.

Patient 1: the widest gap in this article

Inputs: a 50-year-old man, serum creatinine 0.9 mg/dL, not Black.

• MDRD: 175 × 0.9^-1.154 × 50^-0.203. Working through it: 0.9^-1.154 ≈ 1.1293 (since ln(0.9) ≈ -0.10536, times -1.154 ≈ 0.12155, e^0.12155 ≈ 1.1293), and 50^-0.203 ≈ 0.4521 (ln(50) ≈ 3.91202, times -0.203 ≈ -0.79414, e^-0.79414 ≈ 0.4521). Multiplying: 175 × 1.1293 × 0.4521 ≈ 89.3 mL/min/1.73m².
• CKD-EPI (2021): SCr ÷ kappa = 0.9 ÷ 0.9 = 1.0 exactly, so both the min and max terms equal 1^anything = 1. That leaves 142 × 0.9938⁵⁰ × 1 ≈ 142 × 0.73274 ≈ 104.0 mL/min/1.73m².
• Mayo Quadratic: 5.249 ÷ 0.9 ≈ 5.8322; 2.114 ÷ 0.81 ≈ 2.6099; 0.00686 × 50 = 0.343. Summing: 1.911 + 5.8322 − 2.6099 − 0.343 ≈ 4.7903. exp(4.7903) ≈ 120.3 mL/min/1.73m².

Three formulas, one patient, one lab value: 89.3, 104.0, and 120.3. That is a 31-point spread, and it is large enough to move this one patient from KDIGO stage G2 (mildly decreased, under MDRD) all the way to a comfortable G1 (normal or high, under Mayo Quadratic), depending purely on which formula was used to read the exact same blood test.

Patient 2: a smaller but still meaningful gap

Inputs: a 65-year-old woman, serum creatinine 1.1 mg/dL, not Black.

Working through the same three formulas: MDRD gives approximately 49.8, CKD-EPI 2021 gives approximately 55.8, and Mayo Quadratic gives approximately 72.6 mL/min/1.73m². The spread here, roughly 23 points between the lowest and highest, is narrower than Patient 1's 31-point gap but still large enough to shift the clinical picture meaningfully: MDRD and CKD-EPI both land in stage G3a (mildly to moderately decreased), while Mayo Quadratic lands in G2 (mildly decreased), a one-stage difference purely from formula choice.

Patient 3: moving further from normal, the gap narrows further

Inputs: a 35-year-old man, serum creatinine 1.3 mg/dL, not Black.

Here MDRD gives approximately 62.8, CKD-EPI 2021 gives approximately 73.5, and Mayo Quadratic gives approximately 86.3 mL/min/1.73m². The spread is about 24 points, similar to Patient 2, though this patient's higher formulas still land in stage G1 to G2 territory while MDRD sits in G2.

Patient 4: at meaningfully reduced kidney function, the three formulas nearly agree

Inputs: a 70-year-old man, serum creatinine 2.0 mg/dL, not Black, representing a patient with clearly, meaningfully reduced kidney function rather than normal or near-normal function.

MDRD gives approximately 33.2, CKD-EPI 2021 gives approximately 35.2, and Mayo Quadratic gives approximately 34.0 mL/min/1.73m². The spread has collapsed to just 2 points, all three formulas landing solidly in the same stage, G3b (moderately to severely decreased). This is the single most important pattern in this entire comparison, and it is worth stating plainly: the disagreement between MDRD, CKD-EPI and Mayo Quadratic is not a fixed, constant gap that applies everywhere; it is concentrated almost entirely in the normal-to-mildly-reduced range, and it effectively disappears once true kidney function is more clearly impaired.

Patient 5: a young, healthy patient, and a different formula comes out on top

Inputs: a 25-year-old woman, serum creatinine 0.7 mg/dL, not Black, representing a young, healthy patient near the low end of a normal creatinine range.

MDRD gives approximately 102.0, CKD-EPI 2021 gives approximately 123.0, and Mayo Quadratic gives approximately 120.6 mL/min/1.73m². Two things are worth noticing here that the first four patients did not show as clearly. First, the gap between MDRD and the other two is still large, 21 points between MDRD and CKD-EPI, consistent with the same underestimation-at-higher-GFR pattern seen throughout this article. Second, and this is the new observation, CKD-EPI and Mayo Quadratic are now the two closest to each other, effectively tied, rather than Mayo Quadratic sitting clearly highest the way it did for Patients 1 through 3. The consistent part of this pattern is that MDRD trails the other two whenever true GFR is high; which of CKD-EPI or Mayo Quadratic comes out on top of the other two is not fixed, and depends on the specific age, sex and creatinine combination.

Why the gap shrinks as kidney function declines

The explanation traces directly back to how each formula was built. MDRD was derived from the original Modification of Diet in Renal Disease study, whose participants were overwhelmingly patients who already had chronic kidney disease with a measurably reduced GFR; very few had anything close to normal kidney function. A formula statistically fitted almost entirely to impaired-range data naturally performs best in that same impaired range, and drifts further from the truth the further a patient's real GFR sits above the population MDRD was actually trained on. That is exactly the underestimation bias visible in Patient 1 and Patient 2 above, and exactly why it is smallest in Patient 4, whose GFR sits closest to the range MDRD's own source data actually covered.

CKD-EPI was developed afterward specifically to fix this: its source population deliberately included a much larger share of people with normal or near-normal kidney function alongside patients with reduced function, spreading its accuracy more evenly across the whole range rather than concentrating it at the low end the way MDRD does. Mayo Quadratic took a narrower, more surgical approach to the same problem, built and validated specifically to correct MDRD's normal-range underestimation, which is visible directly in the numbers above: Mayo Quadratic produces the highest result in every one of the first three patients, exactly the correction its authors designed it to make, and it converges with the other two only once the true GFR drops into range MDRD was already reasonably accurate for.

How nephrology research actually measures which formula is "better"

It is worth being precise about what "more accurate" means in this context, since none of these formulas can simply be checked against a known correct answer in ordinary clinical practice. Validation studies compare each formula's estimate against a directly measured GFR, usually from inulin clearance or a similar gold-standard clearance study conducted in a research setting, across a large group of patients, then report how close the estimates land using a few standard metrics. One commonly used metric in this literature is called P30 accuracy, the percentage of a formula's estimates that fall within 30% of the patient's actual, measured GFR; a formula with a higher P30 percentage is producing fewer estimates that are wildly off, even if its average bias in one direction or another is similar to a competing formula's. CKD-EPI has generally outperformed MDRD on P30 accuracy across a wide range of GFR values in the studies that established it as the newer standard, which is the more rigorous version of the same underestimation pattern this article's five patients demonstrate by hand.

Formula age matters too, not just formula design

One more factor separates these three formulas that is easy to overlook: how recently each one has actually been revisited. MDRD's core equation has remained effectively unchanged since its 2006 IDMS-traceable recalibration, now two decades old. Mayo Quadratic, published in 2004, has likewise not been formally updated since. CKD-EPI, by contrast, has been revised within the last several years, its 2021 update specifically responding to a documented equity problem identified in its 2009 predecessor, a level of active reconsideration neither of the other two formulas has received. This is not a claim that older automatically means worse; the underlying statistical fitting behind MDRD and Mayo Quadratic has not changed because there was no equivalent finding requiring it to change. It is simply another reason CKD-EPI has become the formula most current clinical guidelines point to by default, alongside its stronger P30 performance: it is the one that has most recently been checked against new evidence and revised in response.

What this means if you are reading your own lab result

If your own eGFR came from just one of these three formulas, and you are wondering how much it might change under a different one, this article's pattern gives a genuinely useful rule of thumb: the closer your result already is to the normal range (above roughly 60 to 90 mL/min/1.73m²), the more a different formula could plausibly move the number, potentially by 20 to 30 points in either direction. The lower your result already is, meaningfully below 45 or so, the less a different formula is likely to change the picture; all three tend to agree closely once kidney function is clearly, measurably reduced. Neither observation is a reason to distrust any single result. It is a reason to treat one eGFR reading, from any one formula, as a data point rather than a verdict, exactly as this site's own GFR calculator page's repeated disclaimer states, and to let a doctor interpret a reading, or a trend across several readings, in the context of your full history rather than a single number in isolation.

A note on the race question, briefly

MDRD's published formula, and the 2009 version of CKD-EPI that has since been replaced, both included a coefficient that raised the calculated result for patients identified as Black. That coefficient's removal from CKD-EPI in 2021 followed a specific, documented finding: the assumption behind it, that Black patients as a group had systematically higher creatinine at any given true GFR, was not well supported, and applying it risked making kidney disease look less severe on paper than it actually was, a pattern linked to delayed referrals and delayed transplant listing for some patients. This article's own worked examples above all use "not Black" for every patient specifically to keep MDRD, CKD-EPI, and Mayo Quadratic on the most directly comparable footing; MDRD's coefficient is shown on this site's own calculator purely as a reference figure against MDRD's original, published form, never as a recommendation to apply it.

Putting normal GFR numbers next to the formula comparison

It helps to put this article's formula-to-formula comparison next to the plain question most readers actually start with: what is normal GFR in the first place. Commonly cited GFR reference range figures put normal GFR for adults under 40 at roughly 100 to 130 mL/min/1.73m² for men and roughly 90 to 120 mL/min/1.73m² for women, with gfr normal ranges declining gradually by age afterward even in people with healthy kidneys; a widely reproduced gfr average by decade runs from around 116 in the 20s down to around 75 by 70 and older. Normal gfr by age and gender, in other words, is not one fixed number, and neither is a normal gfr value for any specific formula, which is exactly the point this article's five patients demonstrate directly: the same egfr normal range boundary can be crossed or not crossed purely by which of MDRD, CKD-EPI, or Mayo Quadratic produced the number being checked against it.

This also connects back to why creatinine clearance vs GFR is worth distinguishing clearly. None of MDRD, CKD-EPI, or Mayo Quadratic actually measures glomerular filtration directly; all three estimate it from creatinine, age, and sex, aiming at true GFR itself rather than at creatinine clearance, which research has shown runs about 10 to 20% above true GFR due to a small amount of creatinine being actively secreted by the kidney tubules rather than only filtered. A cystatin C with eGFR result, or an eGFR by cystatin C calculation using an entirely different blood marker instead of creatinine, is one increasingly used way clinicians cross-check a creatinine-based estimate when the picture is unclear, precisely because it does not share creatinine's muscle-mass dependency or its small secretion bias.

A short checklist for reading three different numbers off one lab report

• If you want to know how to calculate kidney function or how to compute kidney function from a lab value yourself rather than trusting a single online tool, the conversion formula for creatinine reported in micromol/L rather than mg/dL is to divide by 88.4 before running any of the three equations above.
• Expect the widest disagreement between MDRD, CKD-EPI, and Mayo Quadratic when your GFR rating is already close to normal; expect the three to converge closely once your result is clearly, meaningfully reduced.
• Treat CKD-EPI as the reference figure to lean on by default, since it has the strongest current clinical validation and the most recent update, and treat MDRD and Mayo Quadratic as useful context for understanding the spread, not as competing verdicts.
• A GFR calculator, an eGFR calculator, or informally a CKD calculator that only shows one formula is not wrong, it is simply choosing not to show you the spread this article walks through by hand; running the same numbers through a tool that shows all three, and comparing that against gfr stages chart boundaries, gives a fuller picture than any single number can.

Running your own comparison

Every number in this article came from typing the same three or four inputs into each formula in turn and checking the arithmetic independently. If you want to see your own creatinine, age, and sex run through all three formulas at once, the GFR calculator on this site does exactly that: CKD-EPI 2021 as the large, primary, race-free result, with MDRD and Mayo Quadratic alongside it as reference figures, plus a separate Schwartz-formula calculator for children on its own tab. Whichever number you end up looking at, this article's own pattern is the thing worth remembering: expect more disagreement between formulas the closer your result is to normal, and expect that disagreement to fade the more clearly reduced your kidney function already is.