From triglycerides of 174 mg/dL upward (1.97 mmol/L on the 2.2 divisor), every divisor in the Martin-Hopkins table is 5.1 or more, we found on checking all 180, so in that range a Martin-Hopkins LDL can never come out below a Friedewald LDL from the same three numbers. A calculated LDL is total cholesterol minus HDL cholesterol minus an estimate of the cholesterol carried in VLDL particles. Friedewald and Martin-Hopkins differ only in that estimate; Sampson's equation also re-weights total and HDL cholesterol and adds a constant. On one set of inputs we chose, total cholesterol 3.6 mmol/L, HDL 1.0 and triglycerides 4.0, Friedewald gives 0.78 mmol/L, Martin-Hopkins nearly twice that at 1.50, and Sampson 1.22. None of the three is a measurement of LDL.
Friedewald takes away a fifth of triglycerides for VLDL
Friedewald, Levy and Fredrickson published the method in Clinical Chemistry in 1972, from fasting measurements of all three. LDL, HDL and VLDL are the low-, high- and very-low-density lipoproteins, the particles that carry cholesterol in the blood; triglycerides are the blood fats counted by the third result. In mg/dL the formula is LDL = TC - HDL - TG/5, the fifth of triglycerides standing in for the cholesterol inside VLDL.
Reports in mmol/L need another divisor, because cholesterol and triglyceride molecules weigh different amounts: 1 mmol/L is about 38.67 mg/dL of cholesterol but 88.57 mg/dL of triglyceride. On those factors our arithmetic turns the fifth into triglycerides divided by 2.183. Martin and colleagues credit Friedewald's own team with seeing the weakness: "simply dividing triglyceride values by 5 does not give an accurate estimate" of VLDL cholesterol.
Martin-Hopkins factors in non-HDL cholesterol through a 180-cell table
Martin's team, writing in JAMA in 2013, studied 1,350,908 US lipid profiles from 2009 to 2011, with cholesterol separated by ultracentrifugation. The median ratio of triglycerides to VLDL cholesterol was 5.2, but individual ratios ran from 0.4 to 145. Their method swaps the fixed 5 for a factor read from a table of 30 triglyceride rows by six columns of non-HDL cholesterol, which is total cholesterol minus HDL.
For triglycerides under 400 mg/dL, Martin-Hopkins put 91.7% of profiles in the same guideline category as the measured LDL, against 85.4% for Friedewald. Among results estimated under 70 mg/dL with triglycerides of 200 to 399 mg/dL, the measured LDL was also under 70 in 84.0% of cases for Martin-Hopkins and 40.3% for Friedewald.
Sampson squares the triglyceride term
Sampson and colleagues, in JAMA Cardiology in 2020, fitted their equation to 18,715 β-quantification results from 8,656 NIH Clinical Center patients, collected from 1976 to 1999. β-quantification is the reference laboratory method, which spins off the VLDL in an ultracentrifuge before measuring. In mg/dL the LDL equation reads TC/0.948 - HDL/0.971 - (TG/8.56 + TG*nonHDL/2140 - TG^2/16100) - 9.44.
The squared term carries a minus sign inside the VLDL bracket, and the bracket is itself subtracted, so it raises the LDL estimate. The authors say the equation can be used up to 800 mg/dL of triglycerides, and report a root-mean-square error against β-quantification of 15.2 mg/dL, against 32 for Friedewald.
The paper presents that equation "with the same terms for VLDL-C as Equation 1", its VLDL equation, yet the constants differ: Equation 1 divides triglycerides by 8.59, for one. By our arithmetic, subtracting Equation 1's VLDL from total minus HDL instead moves the answer by up to 0.14 mmol/L across the input combinations we tested.
Once triglycerides reach 1.97 mmol/L, Friedewald subtracts more than Martin-Hopkins
Above that point every Martin-Hopkins factor exceeds 5, so the VLDL it subtracts is smaller than Friedewald's. Below it the table holds factors under 5. With the 2.2 divisor that SGH's page and Malaysia's guideline use, total cholesterol 3.6 mmol/L, HDL 1.0 and triglycerides 1.0 give 2.15 mmol/L by Friedewald and 2.13 by Martin-Hopkins. Raise the triglycerides to 2.0 and Friedewald's 1.69 falls under 1.8 mmol/L (70 mg/dL), while Martin-Hopkins gives 1.90 and Sampson 1.82.
We ran all three over 39,729 combinations in 0.1 mmol/L steps: total cholesterol 3.0 to 8.0, HDL 0.7 to 2.5, triglycerides 0.5 to 4.5. Of the 31,543 where all three give an LDL between 0.5 and 5.0 mmol/L, we took those Friedewald placed under 1.8 mmol/L; Martin-Hopkins placed 7% of them at or above it when triglycerides were 1.0 to 1.4 mmol/L, rising to 58% at 4.0 to 4.4. Those shares move with the ranges we chose and say nothing about people. The newer formulas disagree with each other too: with triglycerides of 2.0 to 4.5 mmol/L, Sampson put 1,191 of 20,133 combinations under 1.8 that Martin-Hopkins put at or over it, and none the other way.
Until 12 October 2026 our LDL calculator worked in mg/dL and computed Friedewald alone. It could print a negative LDL under its lowest band when HDL came close to or above total cholesterol, described Martin-Hopkins as a direct measurement, and carried treatment suggestions attributed to a guideline since retired. The revised version computes all three in either unit, refuses an HDL at or above total cholesterol and assigns no bands.
SGH and Malaysia's guideline divide triglycerides by 2.2
Singapore General Hospital's laboratory test page describes its LDL as "a calculated parameter using the Friedewald equation", divides triglycerides by 2.2 in mmol/L, and calls the result invalid when triglycerides are above 4.5 mmol/L. Malaysia's 2017 Clinical Practice Guideline on dyslipidaemia prints the same formula and limit, and says LDL is usually calculated that way. By our arithmetic, 2.2 in place of 2.183 shifts a result by at most 0.016 mmol/L; across our 39,729 combinations it moved 185 across a line at 1.4, 1.8 or 2.6 mmol/L.
Singapore's ACE clinical guideline on lipid management (December 2023, amended June 2025) draws lines at 1.4 mmol/L (55 mg/dL) for patients with a history of acute coronary syndrome and 1.8 mmol/L (70 mg/dL) for other patients with atherosclerotic cardiovascular disease, under the heading "LDL-C target". It names no method for calculating LDL.
A Health Sciences Authority study (Tan and colleagues, 2023) compared the three equations against reference-method values on 345 datasets from five years of HSA's external quality assessment programme, in which many laboratories test the same serum samples. These were quality-control samples rather than patients. The abstract reports the smallest median discordance for Martin-Hopkins, -0.725%, against -4.094% for Friedewald, and says that in samples with high triglycerides, low HDL and high LDL, Martin-Hopkins misclassified none and Friedewald about half.
Converting units rounds the cut-off for direct LDL slightly low
The 400 mg/dL limit used with Friedewald converts to 4.516 mmol/L, and SGH's 4.5 mmol/L converts back to 398.6 mg/dL, so a laboratory working in mmol/L stops calculating a shade earlier. Above that point SGH's page marks the calculation invalid, and Malaysia's guideline says LDL "will have to be measured directly". A direct LDL is an assay that measures it in the sample.
The 2018 AHA/ACC cholesterol guideline called for a repeat fasting profile when a non-fasting one showed triglycerides of 400 mg/dL (4.5 mmol/L) or more. For adults with LDL under 70 mg/dL (1.8 mmol/L), it said "measurement of direct LDL-C or modified LDL-C estimate is reasonable". The 2026 ACC/AHA dyslipidemia guideline, published on 13 March 2026, US Eastern time, retires and replaces it. The American Heart Association's announcement names no formula for calculating LDL.
View author profile → · Editorial policy
