Somewhere in the prescribing information for the blood thinner you or a relative may be taking, there is an equation. It decides the dose. In some cases it decides whether the drug is given at all.

It was fitted, in 1976, to seven points on a graph. All of them came from men. And the multiplier that adjusts the answer for women — the 0.85 that every calculator on the internet applies, including ours — has no female measurements behind it whatsoever.

None of that is a scandal or a secret. The first author wrote it down himself, sixteen years later, in a short retrospective he was invited to write precisely because the paper had become one of the most-cited in medicine.

What Cockcroft said about his own paper

In 1992, asked to reflect on the 1976 paper for a Citation Classic commentary, Donald Cockcroft described how the equation was built. Cockcroft's account is unusually candid, and it is the primary source for this section.

He had collected 505 pairs of creatinine clearance measurements. All were from men. Of those, 249 pairs were, in his words, "reproducible and arbitrarily assumed to be 'accurate'" — and the remainder, a little over half of everything collected, were set aside.

The next part changes how the equation should be read. The surviving data were "condensed by a statistical trick to 7 data points" — the mean creatinine excretion for each decade of age — and the regression that produced the famous formula was run on those seven averages.

⚠️ This is not a quibble about sample size. The number people cite as the equation's n is wrong. The formula was not fitted to 249 patients; it was fitted to seven points, each an average for a decade of age, with the internal spread already smoothed away.

And in that commentary, Cockcroft prints the formula with no sex term at all.

Where the 0.85 came from

The 1976 abstract describes the equation as applying "in adult males, with 15% less in females."

That is a sentence, not a result. There were no women in the 505 pairs, none in the 249, and none in the seven points the line was drawn through. The 15% was a reasonable inference from what was then known about muscle mass and creatinine production — but it is an assumption carried forward for fifty years inside a formula that otherwise presents itself as empirical.

The distinction between "derived" and "assumed" survives nowhere in the way the number is used today. It appears in drug labels as a bare multiplier, in the same typeface as everything else.

The factor is load-bearing, and it works in both directions

It would be easy to assume a 15% reduction is conservative — a safety margin. For at least one drug, it is the opposite.

Edoxaban's US prescribing information prints the Cockcroft-Gault equation, including (0.85 if female), and carries a limitation of use that inverts most people's intuition: the drug should not be used when kidney function is too good. Above a creatinine clearance of 95 mL/min, the 60 mg dose was less effective than warfarin in the trial population, so the label excludes those patients.

Work a case through the label's own equation. A 60-year-old weighing 70 kg with a serum creatinine of 0.75 mg/dL:

  • On the male form: 103.7 mL/min — above 95, so outside the label.
  • With the 0.85 applied: 88.1 mL/min — below 95, so eligible for the full dose.

For the same patient, creatinine and weight, the factor with no female data behind it creates the entire gap between those two outcomes. It alone decides eligibility, in opposite directions, depending on which side of the threshold the arithmetic lands.

The regulators have moved. The labels have not.

This is not an unnoticed problem inside medicine. FDA finalised a guidance on pharmacokinetics in patients with impaired renal function on 18 March 2024, replacing a 2020 draft. A 2025 consensus paper from a National Kidney Foundation workgroup, published in the American Journal of Health-System Pharmacy, characterises that guidance as recommending estimated GFR over Cockcroft-Gault, and tells US health systems to transition away from the older equation for drug dosing.

⚠️ We are quoting the kidney workgroup's reading of the FDA guidance, not the guidance itself. Its existence, its finality and its date we confirmed in the Federal Register; the document's own wording we could not retrieve, and we are not going to put words in a regulator's mouth on the strength of a summary. If that distinction matters for your purposes, read the guidance directly.

Meanwhile the edoxaban label still prints the 1976 equation, and still uses its output to exclude patients. That gap — between what the guidance recommends and what the labels instruct — is why the old equation cannot simply be retired. A prescriber following a label is required to use the equation that label specifies.

eGFR is not a drop-in replacement

If you are tempted to reach for eGFR instead, there is a trap in the units that the kidney consensus makes explicit and almost no consumer calculator mentions.

eGFR is reported per 1.73 m² of body surface area. Cockcroft-Gault is not. One is normalised to a standard-sized body; the other gives an absolute clearance for the actual body in front of you. Substituting one for the other without de-indexing by the patient's own body surface area introduces an error in whichever direction that patient differs from average — and it is largest for exactly the people whose dosing is most delicate, the very small and the very large.

Our eGFR calculator notes the normalisation. Doing the de-indexing needs the patient's actual surface area, which our body surface area calculator will give you.

What to do with this

Nothing, if you are a patient — and that is not a brush-off. Do not recalculate your own clearance and conclude anything about your dose. The equation's weaknesses do not tell you which direction your own number is wrong in, and a label-driven decision made by a prescriber with your actual creatinine history is not improved by an estimate made at home.

What this is useful for is calibration. When a number sits at a threshold — 95, or 30, or 15 — knowing that it came from a seven-point regression on men should change how much weight you put on the third significant figure. A clearance of 96 and a clearance of 94 are not meaningfully different measurements; they are the same estimate landing either side of a line.

Our Cockcroft-Gault calculator implements the equation exactly as the labels print it, 0.85 included, because reproducing what the label says is the only useful thing a calculator can do here. This guide provides the provenance that will not fit in a footnote beneath it.

Where this comes from, and what will date it

The 505 pairs, the 249 kept "arbitrarily assumed to be accurate", the seven-point condensation and the sex-free formula are all from Cockcroft's own 1992 Citation Classic commentary. The "adult males, with 15% less in females" wording is from the 1976 Nephron abstract. The edoxaban equation and the "do not use above 95 mL/min" limitation are from the current US label. The two worked figures are arithmetic on that label's own printed equation, and you can reproduce them.

What we could not verify, and have therefore attributed rather than asserted: the FDA guidance's exact recommendation wording. What we did not examine: whether the other direct oral anticoagulants specify the same equation, and whether the shift to standardised creatinine assays since 1976 has introduced a bias of its own. Both are widely claimed and neither is checked here.

This guide dates when a label changes, not when a guideline does — and labels move slowly, which is the whole reason a 1976 regression on seven points is still deciding doses in 2026.