Lipid and Cardiovascular Markers

Total cholesterol

All the cholesterol carried in blood, across every particle type. It is the least informative of the cholesterol numbers, because it lumps together particles that raise risk and particles that do not, and the 2 pull in opposite directions.

It is still used because it feeds most cardiovascular risk calculators, but the components below carry the actual information. The UK calculator, QRISK3, does not use total cholesterol on its own at all. It uses the ratio of total cholesterol to HDL, which is covered in its own entry below.

Markedly raised cholesterol from a young age, especially with a family history of early heart disease, can indicate familial hypercholesterolaemia, an inherited condition that is common, treatable and substantially underdiagnosed. How the liver handles cholesterol, and what goes wrong in this condition, is in Cholesterol and the lipid system, in the Body Guide.

The word underdiagnosed is doing a lot of work there, so here are the figures. NHS England puts the UK prevalence at about 1 in 250, which is roughly 220,000 people, and puts the proportion currently identified at fewer than 8%. NICE gives a wider figure, between 1 in 250 and 1 in 500. The NHS Long Term Plan set a target of finding 25% of them within 5 years.

Where the numbers come from

NICE arranges specialist assessment where total cholesterol is above 9.0 millimoles per litre, or where non-HDL cholesterol is above 7.5, even when there is no first-degree family history of premature coronary heart disease. That is from its 2023 guideline on cardiovascular risk assessment and lipid modification, which replaced the 2014 version.

Separately, the Simon Broome criteria, which NICE sets out in its familial hypercholesterolaemia guideline, take total cholesterol above 7.5 millimoles per litre (290 milligrams per decilitre) in an adult as the lab part of the criteria. The lab number alone is not the criterion. It also requires either a family history of early heart disease or the physical signs, or a confirmed genetic diagnosis.

Those cholesterol levels alongside tendon xanthomata, which are cholesterol deposits in the tendons, either in the person or in a first or second-degree relative, or alongside genetic evidence of one of the known mutations, give definite familial hypercholesterolaemia. The same cholesterol levels alongside a family history of a heart attack before 50 in a second-degree relative or before 60 in a first-degree relative give possible familial hypercholesterolaemia. The 2 are not the same finding, and the split between them is the substance of the criteria.

Units

The UK reports cholesterol in millimoles per litre, the US in milligrams per decilitre. Divide milligrams per decilitre by about 38.7.

That conversion applies to cholesterol and not to triglycerides, which convert by about 88.5. A figure carried from one to the other is out by more than double, so check which marker a number belongs to before converting it.

HDL cholesterol

Cholesterol carried in high-density lipoprotein particles, which transport it away from tissue and back to the liver. What those particles actually do is in Cholesterol and the lipid system, in the Body Guide.

Higher is not straightforwardly better. The old framing of HDL as simply protective has weakened considerably. Trials of drugs that raise HDL have not delivered the reduction in cardiovascular events that was expected from the observational data, and very high levels are associated with higher mortality. It is best read as part of a pattern instead of as a score to maximise.

Where the numbers come from

No professional body publishes a target for HDL cholesterol, meaning a number to raise it to. HEART UK publishes a desirable level, above 1.0 millimoles per litre in a man and above 1.2 in a woman, and that marks out a low reading rather than something to aim past.

The reason there is no target has been tested repeatedly, and it comes out the same way each time. HDL is the marker most often misread in the optimisation space.

The observational association is real and robust. Across large population studies a higher HDL cholesterol tracks with lower cardiovascular risk, and it does so even between people with the same LDL. It is an input to the cardiovascular risk calculators used in UK practice for exactly that reason, though it enters QRISK3 as part of the total cholesterol to HDL ratio and not as a figure in its own right.

What has not survived is the assumption that raising the number therefore lowers the risk. Across the 3 drug classes that reliably raise HDL cholesterol, which are niacin, the fibrates and the cholesteryl ester transfer protein inhibitors, every one of them raised it, and none reduced all-cause mortality, coronary heart disease mortality or stroke in people already taking a statin.

That last clause is load-bearing. Niacin did reduce non-fatal heart attacks, by about a third, in the trials that ran before statins existed. Added on top of a statin it does nothing, and the difference between those 2 findings is itself statistically solid.

The individual results make the point harder. One drug raised HDL cholesterol by 72% while lowering LDL by a quarter, and increased both cardiovascular events and death from any cause. Blood pressure rose by about 5 millimetres of mercury on it and aldosterone rose too, so an effect unrelated to HDL is the likely explanation, though the mechanism itself could not be ruled out. Another more than doubled HDL cholesterol with no benefit at all.

And adding niacin to statin treatment specifically in people with low HDL, the group most likely to benefit, made no difference whatever to the event rate. A separate trial of niacin in people with established vascular disease also found no benefit, and found excess diabetes, infection, bleeding and a more than 4-fold rise in muscle damage.

There is one exception, and it cuts the other way. The fourth cholesteryl ester transfer protein inhibitor to reach a large outcome trial did reduce coronary events, by about 9%, while roughly doubling HDL cholesterol. It showed no mortality benefit, it accumulates in fat tissue and is cleared slowly, and it was never marketed. The part that matters is why it worked. It also lowered non-HDL cholesterol by 18%, and the genetic evidence indicates the benefit tracked that fall in the harmful particles and not the rise in HDL. When the same genetic analysis pairs a variant of that kind with a statin-like variant, the particle reduction is cancelled out and so is the benefit, despite identical LDL lowering.

So the strict version of the claim has to be narrowed. It is not that no HDL-raising drug has ever reduced an event. It is that where one did, the benefit followed the particles it lowered.

The genetic evidence says the same thing more directly. Inherited variants that raise HDL cholesterol do not lower the risk of a heart attack. One variant raises it by an amount that should cut heart attack risk by 13%, and the measured effect is nil. A score built from 14 such variants shows nothing, in a method that correctly reproduces the effect of LDL when pointed at LDL.

Where the association actually flattens, and where it turns

These are 2 different points and they are usually run together.

The association with coronary events stops improving somewhere around 1.5 millimoles per litre. Above about 2.3, European guidance advises not using HDL as a risk measure at all.

Beyond that, mortality starts to rise again. The lowest all-cause mortality in the Copenhagen population studies sat at about 1.9 millimoles per litre in men and 2.4 in women, with risk climbing above those levels, roughly doubling in men above 3.0.

Two things keep that from meaning what it looks like. The excess is mostly not cardiovascular. In an older cohort, men above 2.07 millimoles per litre had all-cause mortality about 60% higher while their cardiovascular mortality was unchanged, and there was no association in women at all. And it does not appear to be the HDL doing it, because the association had no relationship to the inherited component of a person's HDL level.

The honest reading is that HDL cholesterol is a marker of risk and not a lever. A low value is a real signal and worth paying attention to. A high one is reassuring in the way a good number in a population study is reassuring. A very high one is associated with worse outcomes for reasons that are probably not about the HDL. And there is no evidence that pushing the number higher achieves anything.

LDL cholesterol

Cholesterol carried in low-density lipoprotein particles, and the lipid measure most directly implicated in arterial disease. How those particles get into an artery wall and what happens next is in Heart and blood vessels, in the Body Guide.

On most panels it is calculated instead of measured, using an equation that takes total cholesterol, subtracts HDL, and then subtracts an estimate of the cholesterol in triglyceride-rich particles derived from the triglyceride figure itself.

That subtraction is why triglycerides matter here, and the direction usually gets stated backwards. Rising triglycerides make the subtracted estimate larger, so they push the calculated LDL down, not up. On a non-fasted sample, triglycerides rise by about 0.3 millimoles per litre and calculated LDL falls by about 0.2.

It remains central to guidelines, and the 2026 US dyslipidaemia guideline is explicit that it is not the only marker that matters.

Which equation your lab uses, and why it changed

The oldest and still most common equation, Friedewald, dates from 1972 and is reliable only below a triglyceride of 4.5 millimoles per litre. NICE treats it as needing a fasted sample, and that 4.5 figure is the one NICE quotes.

UK practice has moved. HEART UK and the Association for Laboratory Medicine, in a joint statement published in 2025, now recommend the Sampson equation instead, which holds up to a triglyceride of 9.0 and does not need fasting. NHS labs are switching over, and NICE itself now hedges, saying LDL results may not be reported where triglycerides are above 4.5 millimoles per litre or above 9, depending on the formula the local lab uses.

A third equation, Martin-Hopkins, is more accurate than Friedewald within the old range, particularly at low LDL levels, and in its original form it does not extend the triglyceride range. An extended version has since been published that reaches 9.0, the same span as Sampson. The 2026 US guideline prefers both Martin-Hopkins and Sampson over Friedewald, for greater accuracy across the whole span of LDL and triglyceride levels, and says so most firmly where triglycerides are above 1.7 millimoles per litre or LDL is below 1.8.

Which one produced your number depends on your lab, and your report will not usually say.

How much a non-fasted sample actually changes

Less than most people assume, and the size of the shift decides whether a non-fasted result is usable.

After normal food, total cholesterol falls by about 0.2 millimoles per litre, LDL by about 0.2, non-HDL by about 0.2, and triglycerides rise by about 0.3. HDL cholesterol, apolipoprotein B, apolipoprotein A1 and lipoprotein(a) are not affected by whether you have eaten.

So a non-fasted LDL is shifted, slightly, downward. It is not rendered meaningless, and the markers that do not move are the reason several guidelines have stopped requiring a fast. The 2026 US guideline puts it plainly, that a non-fasting profile suffices in most cases unless someone is known to have high triglycerides.

Where the numbers come from

The 2026 US dyslipidaemia guideline restored explicit LDL cholesterol goals, which US guidance had not carried since 2013. American guidance works in milligrams per decilitre, and the goals run below 100 (2.6 millimoles per litre) where 10-year risk is under 10%, below 70 (1.8) where it is 10% or more and in established cardiovascular disease that is not very high risk, and below 55 (1.4) in established disease at very high risk.

It sets paired non-HDL goals alongside each of those, at below 130, 100 and 85 milligrams per decilitre, which are 3.4, 2.6 and 2.2 millimoles per litre. Those are the more transferable figures for a UK reader, because UK guidance works in non-HDL.

These are treatment goals for someone already being managed, not thresholds that apply to everyone. They sit in a table without recommendation grades attached, so they carry less formal weight than a numbered recommendation does.

UK guidance commits to its own figure and it is easy to miss. NICE's 2023 guideline sets a secondary prevention target of an LDL cholesterol at or below 2.0 millimoles per litre, or a non-HDL cholesterol at or below 2.6. That is the number a UK reader's GP is working to.

The Joint British Societies recommendations, which predate it, advise a non-HDL cholesterol below 2.5, corresponding roughly to an LDL of 1.8. NHS England's national lipid pathway carries those tighter figures alongside NICE's, and it notes where NICE's guideline does not cover the same ground. So there are 2 live UK numbers and they disagree slightly, and neither is the single answer.

Separately, the Simon Broome criteria take LDL cholesterol above 4.9 millimoles per litre (190 milligrams per decilitre) in an adult as the lab part of the criteria for familial hypercholesterolaemia, alongside the clinical features described under total cholesterol. The 2026 US guideline now allows an apolipoprotein B at or above 140 milligrams per decilitre to define the same territory on its own, which is the first time that marker has carried a diagnostic threshold. NICE's familial hypercholesterolaemia guideline puts adult LDL cholesterol above 13 millimoles per litre in the territory of the homozygous form.

Non-HDL cholesterol

Total cholesterol minus HDL, which leaves all the cholesterol carried in potentially artery-damaging particles.

Its practical advantage is that it needs no estimate derived from triglycerides. So it does not carry the equation problem described in the LDL entry above. It does move a little with food, by about the same 0.2 millimoles per litre that total cholesterol falls, because HDL itself does not move at all. European consensus on non-fasting samples treats a shift of that size as not clinically significant.

It is also at least as good a predictor as LDL, and it costs nothing extra, because it is arithmetic on 2 numbers the panel already produces. Where the 2 have been compared directly in people on statin treatment, non-HDL was the better of the 2.

Where the numbers come from

NICE commits to several figures for this marker.

Specialist assessment is arranged where non-HDL cholesterol is above 7.5 millimoles per litre, or where total cholesterol is above 9.0, even without a first-degree family history of premature coronary heart disease.

For primary prevention, NICE aims for a reduction in non-HDL cholesterol of more than 40%, and measures a full lipid profile 2 to 3 months after starting or changing lipid-lowering treatment. For secondary prevention it sets an absolute target, at or below 2.6 millimoles per litre, or an LDL at or below 2.0.

The Joint British Societies recommendations advise a non-HDL cholesterol below 2.5, and NHS England's national lipid pathway carries that figure alongside NICE's, noting where NICE's guideline does not cover the same ground.

The 2026 US guideline sets non-HDL goals too, at below 3.4, 2.6 and 2.2 millimoles per litre by risk category, alongside its LDL goals.

The 40% figure is a different kind of number from the rest. It is a target expressed as a change from your own starting point instead of as an absolute level, so it depends on what your untreated number was.

One point of precision about which guidelines prefer it. UK guidance genuinely prefers non-HDL to LDL for routine use, and NICE says why, which is that LDL needs a calculation and a fasted sample while non-HDL needs neither. Elsewhere the position is weaker than it is sometimes made to sound. The 2026 US guideline sets goals for both and treats them as equals. European guidance keeps LDL as the main focus. HEART UK's advice is to report both.

HDL as a percentage of total cholesterol

A ratio instead of a measurement, giving a sense of the balance between particle types instead of absolute quantities.

Ratios can be useful for tracking change over time, but they hide the underlying numbers. The same ratio can arise from very different lipid profiles.

Where the numbers come from

There is no published threshold for this figure, and none for the other form of it either.

Every cardiovascular risk calculator that uses this relationship expresses it the other way up, as the total cholesterol to HDL ratio. QRISK3, which is the calculator NICE specifies, takes that ratio as a continuous input and does not compare it against any cut-off. The European calculator, SCORE2, does not use a ratio at all and takes total cholesterol and HDL as separate figures. The American calculator changed in 2026, from the pooled cohort equations to one called PREVENT, which covers adults aged 30 to 79 and estimates risk over 30 years as well as 10. None of the 3 uses the percentage form.

The 2 forms are the same information. Divide 100 by the percentage to get the ratio. So an HDL of 25% of total cholesterol is a ratio of 4.0, 20% is a ratio of 5.0, and 33% is a ratio of 3.0.

That conversion matters, because without it a number on your report cannot be compared against anything published anywhere.

Figures like above 4, above 5 or above 6 circulate widely for the ratio form, and they trace back to research from the Framingham era rather than to any current guideline. HEART UK's position is that the ratio is needed for calculating risk with QRISK and should otherwise be read with caution, because it can look acceptable while the numbers underneath it are not.

Like the unsaturated iron binding capacity in the iron group and globulin in the liver group, it is arithmetic from 2 numbers already on your report instead of a separate measurement.

Triglycerides

The main form in which fat is transported and stored. Raised levels are strongly associated with insulin resistance and metabolic dysfunction. How fat is moved around the body and where it ends up is in Cholesterol and the lipid system, in the Body Guide.

This is the lipid marker most affected by food, and it is the only lipid marker that meaningfully moves after a meal. The rise is about 0.3 millimoles per litre after normal eating, which is smaller than most people expect but enough to matter, because the calculated LDL is derived partly from this number and moves in the opposite direction as a result.

They also rise with alcohol, sometimes dramatically after a heavy night.

Where the numbers come from

NICE sets out 3 bands.

Above 20 millimoles per litre, where it is not explained by alcohol intake or poor blood sugar control, NICE refers for urgent specialist review. NICE does not say why in that recommendation, and the concern at that level is pancreatitis instead of cardiovascular risk. The 2025 HEART UK statement endorses that 20 figure as the point for an urgent alert, and it reports a lower threshold from European lab medicine guidance, which suggests flagging a triglyceride above 10 as carrying a high risk of pancreatitis.

Between 10 and 20, NICE repeats the measurement as a fasting test after an interval of 5 days but within 2 weeks, reviews possible secondary causes, and seeks specialist advice if it remains above 10.

Between 4.5 and 9.9, NICE says cardiovascular risk may be underestimated by the standard risk assessment tools at these levels. The same recommendation adds 2 things. Optimise the other cardiovascular risk factors that are present, and seek specialist advice if non-HDL cholesterol is above 7.5.

Below 4.5 there is no committed figure, and the population reference range is assay-dependent. That 4.5 figure appears again under LDL cholesterol, because it is also the point above which the oldest LDL calculation stops being reliable.

One further figure belongs with these, because it is the one the apolipoprotein B entry turns on. The 2026 US guideline uses a triglyceride at or above 1.7 millimoles per litre as the level at which LDL cholesterol starts to understate the particle burden.

Whether lowering them does anything, which is the question the bands do not answer

Triglycerides themselves do not appear to be what causes the damage. The genetic evidence indicates that the cardiovascular risk carried by triglyceride-rich particles collapses onto the number of those particles, which is what apolipoprotein B measures. Once particle number is accounted for, the association between triglycerides and heart disease goes to nothing.

That has a practical consequence. Lowering triglycerides reduces events where it also lowers particle number, and not otherwise. A drug that cut triglycerides by about a quarter without lowering apolipoprotein B did not reduce cardiovascular events at all.

So a raised triglyceride is a real signal, and it is a signal about the particles and the metabolic state behind them instead of about the fat molecules themselves. That is the same structural point the HDL entry above makes about a different number.

Remnant cholesterol

One more term comes up in this area, and it is calculable from a standard panel. Remnant cholesterol is the cholesterol carried in triglyceride-rich particles and their remnants, and it is total cholesterol minus HDL minus LDL, where the LDL has been measured directly. Where the LDL was calculated instead, that subtraction is partly circular, because the calculation already derived part of the answer from the triglyceride figure. It is not a separate measurement and it is not on any UK report, but it is the fraction the triglyceride number is standing in for.

Units

The UK reports triglycerides in millimoles per litre, the US in milligrams per decilitre. Divide milligrams per decilitre by about 88.5. That is a different factor from the one used for cholesterol, which is about 38.7, and the 2 are not interchangeable.

Apolipoprotein B

ApoB

One apolipoprotein B molecule sits on every artery-damaging lipid particle and stays with it for the whole of its life in the circulation. So measuring apolipoprotein B counts the particles directly instead of measuring the cholesterol inside them.

Two forms exist. The longer one sits on LDL, on intermediate and very low density particles and on lipoprotein(a). A shorter one sits on the particles made in the gut after a meal. Standard assays measure both, so the result is a total particle count, and the gut particles are less than 1% of a sample even after eating.

Two exceptions to the idea that every one of these particles is equally damaging, and both are named in the UK consensus statement. Lipoprotein(a) adds risk on top of its particle count, which is covered in the lipoprotein(a) entry below. And in one uncommon inherited disorder the remnant particles are abnormally loaded with cholesterol and do more damage per particle.

This is why it has gained ground. Two people can have identical LDL cholesterol while one carries far more particles, and it is particle number that drives arterial damage.

How often the 2 numbers disagree, and why no single figure answers it

Figures like 1 person in 5 circulate for this, and that is the bottom of the published range. No single figure is meaningful on its own.

Published discordance between LDL cholesterol and apolipoprotein B runs from about 20% to about 60% of a population, and where it lands depends on which LDL cut-point you pair with which apolipoprotein B cut-point, and on which equation produced the LDL. An apolipoprotein B of 100 milligrams per decilitre and an LDL of 100 do not sit at the same point in the population distribution, so pairing those 2 is not comparing like with like. Pair figures at matched percentiles and discordance looks large. Pair them as they are usually quoted and it looks small.

For the direction that matters here, where LDL is at target and apolipoprotein B is higher than that LDL would suggest, the figure usually quoted is around 1 person in 3.

Which equation produced the LDL changes that answer completely, and this is the part almost nobody carries across. Those discordance figures rest on LDL estimated by Friedewald. The 2026 US guideline states that the Martin-Hopkins method markedly reduces discordance with apolipoprotein B, and that among people whose LDL is below 1.8 or below 2.6 millimoles per litre by Martin-Hopkins, only about 2% and about 1% have an apolipoprotein B above target. So the case for adding apolipoprotein B is strongest against a Friedewald LDL and much weaker against a Martin-Hopkins one, which ties straight back to the equation question in the LDL entry above.

Put without a rate at all, at an LDL cholesterol of 130 milligrams per decilitre, apolipoprotein B across the population runs from about 86 to 109. Discordance is concentrated in metabolic syndrome, type 2 diabetes and obesity, and it is not confined to them, because metabolically healthy people show it too.

How much better it actually is

The causal case is strong and the predictive case is real but modest, and those are different claims.

On causation, the genetic evidence is about as clear as this field gets. The risk carried by LDL and the risk carried by triglyceride-rich particles both collapse onto apolipoprotein B, and once particle number is accounted for the associations of LDL cholesterol and of triglycerides with heart disease disappear. Lowering particle number by a given amount produces the same benefit whether you do it by a mechanism that lowers LDL or by one that lowers triglycerides.

On prediction it is consistently better than LDL cholesterol, and usually but not always better than non-HDL cholesterol. For an individual sitting at a given LDL, the gap is worth having. At an LDL of 130 milligrams per decilitre, 10-year risk was about 7.3% in those with a high apolipoprotein B against 4.0% in those with a low one, which is close to a doubling at the same cholesterol.

Against that, in the largest UK dataset it added almost nothing to a risk model that already contained total and HDL cholesterol. And in people on statin treatment, pooled across trials, non-HDL cholesterol performed slightly better than apolipoprotein B. Anyone describing it as dramatically superior is overstating it. Non-HDL is a close second and wins on cost and availability.

It is directly measured instead of calculated, by an automated immunoassay, and it is internationally standardised, which is unusual and is a real advantage over lipoprotein(a), with international work to tighten that standardisation still running. It is valid non-fasted, and it is not yet available in every UK lab.

One measurement limitation. At very high triglyceride levels the assay is affected by light scattering from the large particles present, so the non-fasted validity holds at ordinary triglyceride levels and degrades at high ones.

Where the numbers come from

Published figures exist, from several bodies, and they do not agree with each other.

European guidance sets secondary goals of below 65, 80 and 100 milligrams per decilitre for very high, high and moderate risk, to be used once LDL targets have been reached. The US National Lipid Association sets 60, 70 and 90 for its own risk tiers. European lab medicine guidance says labs should flag a result at or above 100. American cardiology guidance used above 130 as a factor that raises estimated risk until 2026, when the figure moved to 120, which is set out below.

The 2026 US dyslipidaemia guideline gives the marker a formal role and defines it narrowly. Apolipoprotein B testing can be useful to improve risk assessment and guide treatment once the LDL and non-HDL goals have already been met, and particularly where triglycerides are at or above 1.7 millimoles per litre, in diabetes, or where the achieved LDL is below 1.8. That is a supporting grade and not a recommendation to measure it in everyone, and the pattern it is looking for is most common in established cardiovascular disease, cardiometabolic disease and diabetes.

It does carry figures, and they are lower than the ones usually quoted for it. The guideline pairs an apolipoprotein B below 100 milligrams per decilitre with an LDL below 2.6 millimoles per litre, and in severe inherited high cholesterol it sets an apolipoprotein B goal below 55 once the LDL goal has been reached, to deal with particles left over at target cholesterol. On the other side it lists an apolipoprotein B at or above 120 milligrams per decilitre as a factor that raises estimated risk, in the same line as a persistently raised LDL between 4.1 and 4.9 millimoles per litre or a non-HDL between 4.9 and 5.7, and it uses at or above 140 as one of the 3 ways to define severe hypercholesterolaemia, the others being an LDL at or above 4.9 millimoles per litre and a non-HDL above 5.7. Its full goal table sits in a figure rather than in the text.

The accurate statement is narrower than any of those figures suggest. HEART UK puts it precisely. Apolipoprotein B does not have validated decision thresholds, because the trials these targets are built on used LDL cholesterol as their endpoint and not apolipoprotein B. So published targets exist and trial-validated ones do not, which is a narrower and more defensible thing to say than that there are no numbers.

The question is also live rather than settled. The 2026 US guideline ranks apolipoprotein B below LDL and non-HDL cholesterol while its own text concedes the marker is more accurate than either, and that inconsistency is being argued about in print. The specific gap is that treating to an apolipoprotein B target has not been shown to improve hard outcomes beyond treating to an LDL or non-HDL target, which is a different thing from the marker being less accurate.

The population reference range is assay-dependent.

Apolipoprotein A1

ApoA1

The main structural protein on HDL particles.

It is not a particle count in the way apolipoprotein B is, and the analogy is often drawn too neatly. HDL particles carry between 2 and 4 apolipoprotein A1 molecules depending on their size, and the number goes up as the particle gets bigger. So the total mass cannot be divided by anything to give a particle count, which is precisely what makes apolipoprotein B unusual.

In practice it tracks HDL cholesterol almost exactly, at a correlation of 0.92, so it carries little information that the HDL figure does not already give you.

It is chiefly used as the denominator in the apolipoprotein B to apolipoprotein A1 ratio.

Where the numbers come from

No guideline sets a treatment target for apolipoprotein A1, and no UK body sets a threshold for it. European lab medicine guidance does publish one, which is a flag at or below 1.25 grams per litre on a non-fasting sample. Nothing published sets a threshold or a target for the apolipoprotein B to apolipoprotein A1 ratio. The population reference range is assay-dependent.

One claim often made for that ratio does not hold. It does not predict cardiovascular risk better than the conventional cholesterol ratios. It predicts about the same. Where they have been compared directly, the apolipoprotein ratio and the non-HDL to HDL ratio came out at effectively identical strength.

The asymmetry with apolipoprotein B is the point, and it follows directly from the HDL entry above. Apolipoprotein B carries guideline weight because lowering the particles it counts has repeatedly been shown to reduce events. Apolipoprotein A1 describes the HDL side of the same picture, and on that side nothing has yet been shown to work by changing the number.

Lipoprotein(a)

Lp(a)

An LDL-like particle carrying an additional protein called apolipoprotein(a), attached to it by a single chemical bond, one molecule per particle. It is an independent risk factor for arterial disease and for calcification of the aortic valve, and it adds risk on top of whatever LDL is doing.

Both of those are supported by genetic evidence and not only by association. For the aortic valve the evidence is unusually clean, with each 10-fold higher level carrying roughly 60% more risk of aortic stenosis.

The clotting story, and which part of it has not held up

Apolipoprotein(a) is built from repeated copies of domains that are 75% to 91% identical to those in plasminogen, the protein the body uses to dissolve clots. A single whole-protein similarity figure is often quoted instead, and no such figure exists, because the 2 proteins are built differently.

One fact decides the argument. Apolipoprotein(a) carries a copy of plasminogen's business end, the part that does the cutting, and that copy is broken. A short deletion collapses the active site and the amino acid that would normally be cleaved to switch it on has been replaced by a different one. Apolipoprotein(a) has no cutting activity at all.

In the lab, isolated apolipoprotein(a) does compete with plasminogen and slow clot breakdown. The whole lipoprotein(a) particle does not, and work published in 2025 found the reason, which is that the critical part of the molecule is physically blocked by the lipoprotein it is attached to.

In people, the plasminogen story has not survived. Lowering lipoprotein(a) sharply, in a small early study of an antisense drug, did not measurably change how fast clots dissolved. And the genetic evidence does not support lipoprotein(a) causing venous clots, in the same work where the arterial findings were positive.

So it is the plasminogen explanation specifically that has failed, and not every thrombotic mechanism. The same 2025 work found that once you allow for lipoprotein(a) speeding up how fast a clot forms, there is a small but real prolongation of clot breakdown after all, and that it produces a fibrin structure more resistant to being dissolved. It also found the particle accelerates thrombin generation, which points the other way, towards clot formation instead of clot persistence. That question is open.

The leading alternative explanation, with its own problem stated

Lipoprotein(a) carries a disproportionate share of the oxidised fats in blood, more than 85% of them, and those are among the strongest drivers of inflammation in an artery wall. That is the mechanism most actively pursued.

It has to face the same test as the clotting story, and it does not pass cleanly either. A drug that cut lipoprotein(a) sharply also cut the oxidised fat load it carries, and it did not move either of the standard markers of inflammation. If a null result on clot breakdown counts against the plasminogen story, a null result on inflammatory markers counts against this one. Neither mechanism is settled.

It is almost entirely genetic. Between 70% and more than 90% of the variation between people is genetic, it is inherited co-dominantly, and the spread across a population is more than 1000-fold. It barely moves with exercise or with statins, and the largest datasets find no change on statins at all. Nobody should stop a statin over a lipoprotein(a) result, because the benefit of the statin does not depend on it.

Levels settle at around age 5 and may drift up into adulthood, so stable in adult life is more accurate than stable for life. That is why the guidance is to measure it once instead of tracking it.

What does move it, since barely moves is not the same as does not move

Several things do, and 2 of them are common enough to matter here. Pregnancy roughly doubles it. Hormone replacement therapy after the menopause lowers it by about a quarter.

Beyond those, nephrotic syndrome raises it 3 to 5-fold, peritoneal dialysis and growth hormone treatment roughly double it, an underactive thyroid raises it and treating the thyroid brings it down by 5% to 20%, and a drug that blocks one of the inflammatory signalling molecules lowers it by 30% to 40%.

And diet does move it, by about 15%, in the direction most people would not predict. Replacing saturated fat with carbohydrate raises lipoprotein(a). A lower-carbohydrate, higher-saturated-fat diet lowers it. That is the opposite of what the same change does to LDL. The 2 markers are pulling apart and nobody has shown which one wins, so this is not yet something to act on.

Where the numbers come from

UK guidance grades the risk instead of drawing one line. The HEART UK consensus position classifies 32 to 90 nanomoles per litre as minor risk, 90 to 200 as moderate, 200 to 400 as high, and above 400 as very high. It gives the population percentiles alongside those bands, which say more than the labels do. The 4 bands sit at roughly the 67th to 80th, 80th to 95th, 95th to 99.8th, and above the 99.8th percentile.

Two caveats belong with those numbers and HEART UK states both. The bands came from a Danish population, and HEART UK says openly that further work is needed to establish cut-off values in other ethnicities. There is also a mismatch across the 2 HEART UK documents, in that the later one names the assay the bands were derived on while the assay HEART UK recommends is a different one.

What that guidance says to do about it has 3 limbs. Reduce overall atherosclerotic risk. Control the rest of the lipid picture, with a desirable non-HDL cholesterol below 2.5 millimoles per litre for anyone above 90 nanomoles per litre. And consider lipoprotein apheresis, which filters the particle out of the blood directly and is used in a small number of people with very high levels and progressive disease.

So it is not quite right to say the response is never to attack the lipoprotein(a) itself. For most people it is to reduce everything else that can be reduced, and the direct option exists at the top end.

The American guideline goes further than the UK one and grades it differently. The 2026 US dyslipidaemia guideline recommends measuring lipoprotein(a) at least once in adulthood, at its strongest class of recommendation, and it is the first US clinical practice guideline to say so. Its stated purpose is to identify people with genetically raised levels who may benefit from earlier or more intensive LDL lowering. The US National Lipid Association said the same thing in 2024, in a scientific statement rather than a guideline.

It treats 125 nanomoles per litre (50 milligrams per decilitre) as the level at which lipoprotein(a) becomes a factor that raises estimated risk, associated with about 1.4 times the risk of atherosclerotic cardiovascular disease, and 250 nanomoles per litre (100 milligrams per decilitre) as associated with at least double. It puts 350 nanomoles per litre (150 milligrams per decilitre) at about 3 times the risk, and at or above 430 (180 milligrams per decilitre) it describes the risk as about 4 times higher, which is where it draws the comparison with the inherited cholesterol disorder described under total cholesterol.

European guidance sits between the 2. The 2025 European focused update on lipid disorders treats a level above 50 milligrams per decilitre, which it converts as 105 nanomoles per litre, as something that modifies estimated risk. It grades that as worth considering rather than as something to do, which is a weaker grade than the American one. Europe says consider, America says do, and that is a real disagreement rather than a wording difference.

UK guidance has not moved. NICE's cardiovascular risk guideline does not mention lipoprotein(a) at all. A UK taskforce convened in 2023, which HEART UK is part of, recommends targeted testing instead of universal, in people with a personal or family history of early arterial disease, in first-degree relatives of someone above 200 nanomoles per litre, in inherited lipid disorders, in calcific aortic valve disease, and where estimated risk is borderline. It has asked NICE to include the marker and to get it into the UK risk calculators.

Measuring it once is explicit UK guidance, not an inference. HEART UK says levels need only be measured once, unless a secondary cause is suspected or treatment is started specifically to lower them.

What a high result is actually for

A raised result does not change what you can do about lipoprotein(a) itself, and it does change how aggressively everything else is usually managed. That is not a consolation prize. In the statin trials, a raised lipoprotein(a) predicted the risk that remained after treatment, independently of how far LDL had been brought down.

It also says something about your family. Because the level is largely inherited, first-degree relatives of someone with a raised result are far more likely than average to have one too. Both American and European guidance recommend testing relatives on that basis.

Ethnicity, which matters for a UK reader

Levels differ substantially between ethnic groups. In UK Biobank the median was about 19 nanomoles per litre in White participants, 31 in South Asian, 75 in Black and 16 in Chinese participants. So a South Asian median is roughly 1.6 times the White median and a Black median roughly 4 times it.

The risk carried per unit appears to be similar across those groups, which is the position most experts take, and it means the same threshold broadly means the same thing. That is not unanimous. One study found a lower threshold identified risk only in Black participants and that a commonly used cut-off did not work in Chinese-American participants, and concluded thresholds should be ethnicity-specific.

And there is an unresolved discrepancy specific to a large UK group. The one dedicated study of South Asian people found no association between lipoprotein(a) and early markers of arterial disease, while the far larger UK Biobank analysis, which counted actual events, did find the expected association in South Asian participants. Nobody has reconciled those.

The units are not interchangeable, and that matters more for this marker than for any other lipid number

Lipoprotein(a) particles vary in size between people, because the number of repeated segments in the protein differs, and the assays are sensitive to that. So a result in milligrams per decilitre cannot be converted to nanomoles per litre by a fixed factor.

It is not that no conversion exists. Several are in use and they do not agree. The 2 guidelines above both convert 50 milligrams per decilitre, and the European update makes it 105 nanomoles per litre while the American one makes it 125, which is the level the American guideline treats as raising estimated risk. The same figure in mass units therefore lands either below that threshold or exactly on it, depending on which conversion is applied.

The paired figures in the American guideline sit at roughly 2.3 to 2.5 times the mass figure, and they are offered as a convenience rather than as a conversion to apply to your own result.

Molar units, nanomoles per litre, are now preferred by every body that has published on it. UK practice has not caught up, and most UK labs still report in mass units.

A lipoprotein(a) number quoted without its unit means nothing. If you are comparing your result against something you have read, check which unit it is in first.

Secondary causes

A raised lipoprotein(a) is sometimes secondary to something else. HEART UK lists 6 to look for which are chronic kidney disease, nephrotic syndrome, underactive thyroid, end stage renal failure on dialysis, autoimmune disorders and growth hormone treatment. Pregnancy and hormone replacement therapy, from the list further up, are the 2 most likely to explain an unexpected result in someone otherwise well.

The first drug trial reported in September 2026, and it did not work

On 4 September 2026 the first outcome trial of a drug built to lower lipoprotein(a) directly reported. The drug lowered lipoprotein(a) as intended and did not reduce cardiovascular events against placebo, in people with established cardiovascular disease who were already on treatment that had brought their LDL low. The full results have not yet been presented, so the size of the effect on events is not public.

Further outcome trials of other drugs are running and one of those drugs is taken by mouth rather than injected.

What that changes for a reader is nothing about what to do, and quite a lot about what to expect. The case for measuring lipoprotein(a) rests on it telling you how hard to work on everything else, and that case is untouched. The case for expecting a drug to fix the number itself is weaker than it was before that result.

One open question at the other end

Very low lipoprotein(a) may not be free. There are signals linking it to new-onset type 2 diabetes and to fatty liver disease. That is unresolved, and it is a reason not to write lower is always better before anyone has shown it.