Every other chapter in this guide treats your lipid panel as a fact. This one is about the gap between the number printed on the report and the thing it is trying to measure — because that gap is bigger than almost anyone is told, and more misinterpretation happens here than anywhere else in the whole subject.
Three separate things are going on. Your cholesterol genuinely fluctuates from week to week. The lab has its own small imprecision. And the LDL-C figure on most UK reports was never measured at all — it was calculated from three other numbers using an equation with known failure modes. Each of these is manageable once you know about it. Together, and unrecognised, they are how someone ends up on a treatment decision made on noise.
1. Your cholesterol is not one number, it is a distribution
Measure the same healthy person repeatedly over weeks and the results scatter. This is not lab error; it is real biological variation, driven by diet in the preceding days, illness, alcohol, weight change, exercise, posture during the blood draw, season, and for women the menstrual cycle. The scatter has been measured many times and the figures are consistent.
| Marker | Typical within-person variation | Derived 95% reference change value |
|---|---|---|
| Total cholesterol | about 6% | roughly 18% |
| LDL-C | about 9% | roughly 25% |
| HDL-C | about 7% | roughly 20% |
| Triglycerides | about 20–25% | roughly 60% |
| Lp(a) | about 10% | roughly 28% |
The right-hand column is the one that matters and the one nobody quotes. It is the reference change value — how much two measurements have to differ before you can be reasonably confident something actually changed rather than the same person being sampled twice from the same underlying scatter. It combines the biological variation with the lab's own analytical imprecision, and it is always larger than people expect.
Work through what that means. An LDL-C of 3.5 mmol/L on one occasion is, very roughly, consistent with a true value somewhere around 2.9 to 4.1. A second reading of 3.2 three months later is not evidence that anything improved — it is well inside the range you would get by taking two samples from an unchanged person. And triglycerides, at 20–25% variation, are close to uninterpretable as a single measurement: a fall from 2.0 to 1.5 sounds like a substantial win and is entirely compatible with nothing at all having happened.
2. That LDL number was probably calculated, not measured
Most UK labs do not measure LDL-C directly. They measure total cholesterol, HDL-C and triglycerides, then estimate LDL-C by subtraction. The classic method is the Friedewald equation, published in 1972:
LDL-C = total cholesterol − HDL-C − (triglycerides ÷ 2.2) · in mmol/L
That final term is an estimate of the cholesterol carried in VLDL, and it assumes a fixed ratio between triglycerides and VLDL cholesterol. In a fasting person with ordinary triglycerides, the assumption holds well enough. Outside that, it degrades — and it degrades in a specific, predictable direction: Friedewald tends to under-estimate LDL-C when triglycerides are high, and again when LDL-C itself is low.
Those are precisely the two situations in which the decision is finest. Someone with metabolic syndrome and triglycerides of 3.0 gets an LDL-C that reads lower than the truth, which is reassurance they have not earned. Someone already on a statin and being assessed against a target of 2.0 gets the same downward bias at exactly the point of judging whether treatment is working.
Two better equations exist, and they disagree with Friedewald often
Both replace the fixed divisor with something adaptive.
| Equation | How it differs | Where it is valid |
|---|---|---|
| Friedewald (1972) | Fixed triglyceride divisor. | Fasting, triglycerides below about 2.3 mmol/L. Conventionally treated as invalid above 4.5 mmol/L. |
| Martin–Hopkins (2013) | The divisor adapts to the person's own triglycerides and non-HDL-C, taken from a table of 180 cells. | Endorsed by European laboratory guidance for LDL-C below 1.8 mmol/L and triglycerides 2.0–4.5 mmol/L. An extended version performs best of the three at 4.5–9.0 mmol/L. |
| Sampson–NIH (2020) | Derived on a hypertriglyceridaemic population, with terms for triglycerides and their interaction with non-HDL-C. | Designed to extend to triglycerides around 9.0 mmol/L. |
How much does the choice matter? More than "an equation refinement" suggests. In a large US cohort with established cardiovascular disease and triglycerides under 4.5 mmol/L, Friedewald and Martin–Hopkins disagreed about whether the patient was below an LDL-C of 1.8 mmol/L in about 15% of cases in the US cohorts where this has been measured. That is not a fixed property of the equations — the proportion moves with the population, the triglyceride distribution and the threshold chosen — so read it as evidence that disagreement is common, not as a rate that applies to you. In the triglyceride band 2.0–4.5 mmol/L, disagreement at that same cut-point ran far higher still. Martin–Hopkins consistently returns the higher figure, which means Friedewald consistently reports more people as being at target than are.
None of this is exotic or contested. It is a laboratory choice made upstream of you, usually not stated on the report, and it can move you across a threshold.
What to actually do with this
- Ask which method your lab used, if a borderline LDL-C is about to change a decision. It is a fair question and the lab will know.
- Look at non-HDL cholesterol instead. It is total cholesterol minus HDL-C — pure subtraction, no equation, no assumption, no triglyceride term to go wrong. NICE frames the treatment response in non-HDL for exactly this reason.
- Or measure apoB. It is measured directly rather than calculated, is unaffected by fasting state, and counts particles rather than estimating the cholesterol inside them. That removes the equation — the second problem on this page — but not the first: apoB has its own biological variation, smaller than LDL-C’s but not zero, so the repeat rule still applies.
- If your triglycerides are above about 4.5 mmol/L, a Friedewald-calculated LDL-C should not be relied on. Ask which equation the lab used, since the newer ones are validated considerably further up, or ask for a direct measurement or an apoB.
3. Fasting or not?
UK and European practice has moved to non-fasting lipid panels as the default, and this is sensible: it is far easier to get people tested, and non-fasting samples predict cardiovascular risk at least as well. Total cholesterol, HDL-C, non-HDL-C and apoB barely move with a meal.
Two caveats survive that shift. Triglycerides do rise after eating, typically modestly but sometimes substantially. And because calculated LDL-C has a triglyceride term in it, a non-fasting sample pushes calculated LDL-C down — by around 0.3 mmol/L on average across the common equations in the large outpatient series where this was measured, and by more in some people. That is not large, but it is the same size as the gap between many treatment thresholds.
The practical rule is about consistency rather than fasting as such: if you are comparing two results to judge whether something changed, compare like with like. A fasting baseline against a non-fasting follow-up can manufacture an improvement that is not there.
4. Everything else that moves the number
These are pre-analytical factors — things that happen before the sample reaches the analyser. None is a laboratory fault and all are avoidable.
- Recent acute illness or infection. Lipids fall during and after significant illness, and can stay depressed for weeks. A panel taken during or shortly after an infection can substantially understate the true position.
- After a heart attack. LDL-C falls within the first day or two of an acute coronary event. A panel taken on day three is not a baseline.
- Alcohol in the preceding days. Raises triglycerides, sometimes markedly.
- Prolonged standing or a tight tourniquet. Haemoconcentration can raise all lipid fractions by several percent — small, but it stacks with everything else.
- Pregnancy. Lipids rise substantially and progressively. Routine lipid assessment is not meaningful during pregnancy or immediately after.
- Untreated hypothyroidism. Raises LDL-C by reducing LDL receptor expression. Treating the thyroid can resolve what looked like a lipid problem, which is why TSH belongs in the baseline panel.
- Changing lab. Different analysers and different LDL-C equations. A change between two results might be a change of laboratory.
Putting it together
None of this is a reason to distrust lipid testing, and it is emphatically not a reason to refuse treatment on the grounds that "the numbers are unreliable". The evidence that lowering atherogenic particles reduces events rests on trials of tens of thousands of people, where random measurement scatter averages out. It is at the level of the individual decision, made on one reading, that this matters.
So the discipline is short and worth keeping:
Repeat anything borderline before acting on it
Same lab, same fasting state, a few weeks apart. Two readings beat one by more than most people would guess.
Judge change against the reference change value, not against the threshold
An LDL-C that moves less than about a quarter, or triglycerides that move less than about half, have not demonstrably moved.
Prefer numbers that are measured over numbers that are calculated
Non-HDL-C and apoB avoid the equation entirely. If you can only add one test to a standard panel, apoB is the one that removes the most uncertainty.
The clinician-facing version of this, with the equation detail and the disagreement rates set out for a consultation, is in the clinical pathway.
One ratio, two unit systems, and a factor of 2.3
The triglyceride-to-HDL ratio is the most widely quoted proxy for insulin resistance, and almost every threshold you will find online is American and therefore in mg/dL. A UK panel reports both numbers in mmol/L.
Those are not interchangeable, and the error does not cancel out. Triglycerides convert at about 88.6 mg/dL per mmol/L and HDL-C at about 38.7, so the same person's ratio is roughly 2.3 times larger in American units than in British ones.
| Commonly quoted target | In mg/dL | The same thing in mmol/L |
|---|---|---|
| “Optimal” | under 2 | under 0.87 |
| “Raised” | 2 to 4 | 0.87 to 1.74 |
| “High” | above 4 | above 1.74 |
| “Aim for 1:1” | 1 | about 0.44 |
The mistake runs in the reassuring direction, which is why it matters. A UK reader with TG 1.9 and HDL 1.1 has a ratio of 1.7. Set against an American target of “under 2” that looks comfortably fine. Converted properly it is 3.9 mg/dL units — near the top of the raised band, and about to cross into high.
Check which units a threshold is in before you compare yourself to it. The bands used elsewhere on this site are stated in mmol/L throughout, and are the converted equivalents of the same American figures.
The same trap sits under every imported lipid target, which is why the guidelines comparison converts all of them rather than quoting them as published.