Liver Markers

ALT

Alanine aminotransferase

An enzyme found mainly inside liver cells. When those cells are damaged, ALT leaks into the blood, which makes it the most liver-specific of the routine enzymes. The UK guideline puts it plainly. ALT sits in low concentrations outside the liver, and elevations that have nothing to do with the liver are uncommon.

It rises with fatty liver, alcohol, viral hepatitis and a long list of medications and supplements.

Hard exercise raises it too, because ALT is also present in muscle. A raised ALT in someone who lifts heavily in the days before the test is a very common and entirely benign finding. The effect outlasts the session by a long way in those unaccustomed to heavy lifting, with ALT, AST and creatine kinase still raised sometimes up to 7 days after.

Sustained elevation, once exercise and medication are excluded, most often reflects fatty liver disease, which is very common and closely tied to weight, insulin resistance and alcohol.

The name for that condition changed in 2023

What used to be called non-alcoholic fatty liver disease, or NAFLD, was renamed metabolic dysfunction-associated steatotic liver disease, or MASLD, by a joint effort of the American, European and Latin American liver societies in 2023. The aggressive form, non-alcoholic steatohepatitis or NASH, became metabolic dysfunction-associated steatohepatitis, or MASH. The change was not cosmetic. The old name defined the condition by what it was not, and required alcohol to be excluded before the diagnosis could be made. The new one defines it by what it is, a liver carrying excess fat alongside at least one cardiometabolic feature, and it creates a separate category for people who have both the metabolic drivers and a meaningful alcohol intake, which is a very large group that the old system had nowhere to put.

UK adoption is partial. You will see both sets of names in circulation, and NHS material has not fully caught up, so a report or a leaflet using the older terms is not out of date in any way that matters clinically. It is the same condition.

Fatty liver, ethnicity and why it is worth taking seriously

Early on, fatty liver is fat sitting inside liver cells and nothing else, and at that stage it comes back out with weight loss. What matters is where it can go next, which is inflammation and then scarring, and the scarring is the part that does not reverse. Liver fat also drives insulin resistance rather than only following it, so it pushes blood sugar in the wrong direction while it is there.

Weight is where ethnicity bites. In UK patients whose fatty liver was confirmed on biopsy, South Asian patients carried the same grade and stage of liver injury as White patients at a body mass index about 4 points lower and about 8 years younger. So a weight that reads as unremarkable does not mean the same thing across groups. The World Health Organization sets lower action points of 23 and 27.5 for Asian populations, and those came from diabetes and cardiovascular risk rather than from liver fat, so they are a prompt to look and not a liver threshold.

How much ethnicity adds on its own is smaller than it is usually made to sound. In the largest UK community study of ethnicity and liver disease, excess weight and diabetes shifted the odds far harder than ethnicity did in any direction, and among South Asian groups only Bangladeshi ethnicity came out independently raised once those were allowed for.

Where the numbers come from

Lab upper limits vary from one hospital to another and are commonly around 40 units per litre. The American College of Gastroenterology puts the true healthy upper limit considerably lower, at 29 to 33 units per litre in men and 19 to 25 in women, and says results above that should be assessed.

The reason for that gap is that a reference range can be right by construction and wrong in practice. Reference ranges are built from people with no known illness, on the assumption that they are healthy. Excess liver fat is common and usually silent, so people who had it sat inside the reference population and pulled the ceiling upward. Rebuild the range from people at genuinely low risk, meaning normal body weight, no viral hepatitis, normal glucose and lipids and no medication, and the limit drops to around 30 units per litre for men and 19 for women.

The American Association for the Study of Liver Diseases has said the same thing plainly. The normal values most labs report exceed what is genuinely normal. American endocrinology guidance uses above 30 units per litre as the point for further assessment without splitting by sex, and the same guideline separately states that the upper limit of ALT should be 30 units per litre for men and 19 for women.

There is a UK figure, and almost nobody quotes it

NICE has published one since 2013. Its guideline on chronic hepatitis B defines a normal ALT as below 30 units per litre in men and below 19 in women, which is the lower end of the American range and considerably below what your report will say. It sits inside a hepatitis guideline instead of a general liver one, which is most of the reason it goes unmentioned, but it is a NICE figure and it exists.

The British Society of Gastroenterology, in its 2018 guideline on abnormal liver blood tests, goes at it from a different angle and arrives somewhere similar. It says that the clinical significance of a liver blood test should not be judged by how far outside the range the result sits, and separately that the current upper limit of normal for many of the liver enzymes may be too high, with ALT as its worked example. So the UK position is not that there is no healthy threshold. It is that the published range is probably too generous and that the distance outside it is not what tells you whether it matters.

So a result of 38 will print as normal on almost any UK report and sits above every healthy threshold named above. That is not an error on the report. It is 2 different questions, one about whether this is outside the population range and one about whether this is a level associated with good health, and only the first is what a reference range answers.

The threshold is not settled, and the newest data pushes it up

Everything above points one way, so here is the work that points the other way, because it is recent and it is large.

Other attempts to define a healthy limit land higher. One puts it at 36 units per litre for men and 29 for women, another at 57 and 35, both above the American figures. And AST and GGT predict hard outcomes better than ALT does, which is not what you would expect if ALT were the marker that mattered most.

None of that makes the 30 and 19 figures wrong. It means the exercise of defining a healthy ALT depends heavily on which population you screen, how hard you screen it and which outcome you then measure against, and different groups doing that honestly have landed several units apart. The direction of travel over the last 20 years has been downward, the 2 most recent large attempts went upward, and there is no body that has adjudicated it.

Nobody publishes an ALT threshold adjusted for ethnicity

Whatever the right number turns out to be, it is the same number for everybody. Every published figure is split by sex, sometimes by age or weight, and never by ethnicity. What does differ is how often a raised ALT turns up, which is a statement about how much fatty liver a population is carrying and not about what counts as normal. The same ALT sits against a different background likelihood depending on who is being tested.

A normal ALT does not mean a normal liver

This is the part of the ALT story that matters most and gets said least. Among people with fatty liver confirmed on biopsy who had a normal ALT, around 35% had bridging fibrosis or cirrhosis. That is the high end of the published range and not a settled figure. Later work put advanced fibrosis at 19% and cirrhosis at 7% in the same situation. The enzyme had nothing to say about it.

That is not an argument against testing ALT. It is an argument against reading a normal one as an all-clear, and it is the reason the guidelines have moved towards scoring systems that combine markers instead of watching a single enzyme.

Fibrosis scores, and where UK guidance splits

FIB-4 is the most widely used of those scores. It combines age, ALT, AST and platelet count into a single figure, all of which are on a standard venous panel, and it needs no extra test.

The British Society of Gastroenterology has recommended first-line fibrosis testing with either FIB-4 or the NAFLD Fibrosis Score since 2018, so it offers 2 scores rather than singling this one out. NICE has not. Its 2016 guideline on fatty liver recommends the Enhanced Liver Fibrosis test instead, a separate paid-for panel, and it says explicitly that routine liver blood tests should not be used to assess for advanced fibrosis. Those 2 positions are hard to reconcile, and the NICE one is the older of the 2 by 2 years. The ELF cut-off is itself disputed. NICE still uses 10.51 and has not changed it, while Health Technology Wales recommends moving to 9.8, which is the test manufacturer's own cut-off and the figure the European liver association has also proposed.

FIB-4 also has a real and specific weakness if you are young. In general-population use its ability to discriminate at ages 18 to 34 is essentially nil, and it performs no better than guessing. It was built and validated in populations where fibrosis was common, and age is one of its inputs, so in a young low-prevalence group it does not work. It becomes useful later in life.

Telling muscle from liver

Because ALT is also present in muscle, a raised ALT and AST in someone who trains hard, with a normal GGT and a raised creatine kinase, is a muscle pattern and not a liver one. Creatine kinase is what separates the 2, and without it there is no way to tell.

How old the guidance actually is

The background to this whole section is that the guidance UK labs and GPs work from is not new.

The American College of Gastroenterology guideline on abnormal liver chemistries is from 2017. The British Society of Gastroenterology guideline on abnormal liver blood tests is from 2018. Neither has been replaced. The NICE fatty liver guideline is from 2016 and predates both the MASLD renaming and most of the fibrosis-scoring evidence. There is however, an update in development, due in December 2026. The NICE hepatitis B guideline that carries the only UK healthy ALT figure is from 2013 and was last updated in 2017. The revised classification used for acute pancreatitis, which is where the lipase threshold below comes from, is from 2012.

Meanwhile the newest attempts to define a healthy ALT were published in 2025, the renaming of fatty liver happened in 2023, and the evidence that most changes what a low ALP means was published in 2026. None of that has been through a guideline yet.

What a marker inside its range is worth

One finding applies to every marker here. Values sitting well inside the reference interval, between the 10th and 90th percentiles, are associated with death from any cause for 20 common markers, albumin and ALP among them.

The same work found that all 152 markers it tested, put together, explained an additional 0.8% of the variation in the risk of dying. So the contribution these markers make on top of what is already known is minor.

Both halves are true, and they are not in tension. Within-range variation in these markers is genuinely, reproducibly associated with outcomes at the scale of a population, and it tells an individual very little about their own future. That holds for every within-range association described below.

AST

Aspartate aminotransferase

An enzyme present in the liver, and also in heart and skeletal muscle, smooth muscle, kidney, brain and red blood cells. Less liver-specific than ALT, which is why the 2 are read together.

Heart and liver hold the most AST by concentration, with skeletal muscle behind them, but muscle is a large tissue and a common source of a raised result. The UK guideline lists AST as abundantly present in skeletal, cardiac and smooth muscle, and notes that it may therefore be raised after a heart attack or in muscle inflammation.

The exception runs the other way. In alcohol-related liver disease AST can be the more sensitive of the 2 enzymes, which is the one common setting where the less liver-specific marker is the better one.

The ratio between AST and ALT is often more informative than either alone, and gives a clue as to whether the source is liver or muscle.

Haemolysis, and what it actually does

Of the liver enzymes, AST is the one most distorted by a haemolysed sample, meaning one where red cells have ruptured between the needle and the analyser.

The distortion starts before haemolysis is visible. A measurable change in AST appears at a serum haemoglobin below 0.6 grams per litre, which is a degree of haemolysis described in the original work as mild or almost undetectable by eye. In a badly haemolysed sample, at around 4.5 grams per litre of free haemoglobin, AST reads roughly 2.5 times higher than it should. Where haemolysed samples were compared against a fresh redraw from the same person, the AST difference exceeded what would count as a genuine change in that person.

Two things about this are commonly got wrong, and both matter if you are trying to decide whether to trust a result.

AST is not the worst-affected marker overall. It is the worst-affected of the liver enzymes. Lactate dehydrogenase and potassium are both affected far more, LDH by around 51% and potassium by around 14% per 100 units of haemolysis index. A 2026 study ranked the order as LDH first, then uric acid, then AST, then bilirubin.

Haemolysis does not simply inflate liver markers. It pushes some of them down. Albumin, ALP and GGT all read falsely low in a haemolysed sample, along with chloride, glucose and sodium. Total bilirubin can read falsely low too, because haemoglobin interferes with the colour reaction the assay depends on, although at least one platform gives a falsely high bilirubin instead, so the direction there is method-dependent.

The reason is that only part of the interference is enzyme leaking out of red cells. A large part of it is optical. Haemoglobin absorbs light at the wavelengths these assays measure at, which is why ALP and GGT go down while AST goes up. Because the size and the direction both depend on the analyser, a haemolysed sample is one to redraw and not one to correct. That is the cautious position rather than the one the measurement work itself reaches, since at mild haemolysis the bias in most markers stays within clinically acceptable limits.

Where the numbers come from

No professional body publishes a healthy threshold for AST in the way the American College of Gastroenterology does for ALT. That guideline gives a healthy figure for ALT and for nothing else on the liver panel, directing clinicians to their own lab's upper limits for AST, ALP, GGT, albumin and bilirubin. The UK guideline gives no standalone AST recommendation either. The population reference range is assay-dependent. Any optimal AST you see quoted is somebody's own invention.

The ratio, and why the alcohol rule is close to backwards

The ratio has a long-standing rule of thumb attached to it. Above 2 in someone drinking regularly is said to point towards alcohol-related liver injury, and below 1 to be more typical of liver fat or viral hepatitis. The first half of that is not what the primary evidence says, and the way it is usually repeated is actively misleading.

In heavy drinkers without severe liver disease, the ratio is at or below 1.0 in 64% of them and only exceptionally reaches 2. In people with established alcohol-related cirrhosis, 69% are at or above 2 and only 8% are at or below 1.0. So a ratio above 2 points to advanced alcohol-related liver disease and not to heavy drinking on its own.

So a ratio above 2 marks advanced disease and not drinking. The consequence that matters more runs the other way round. In a heavy drinker, a normal ratio is the expected finding. That is not reassurance, and it is not evidence that the drinking has done no damage.

The second half of the rule holds up better, with one condition. In fatty liver the ratio averages 0.9, against 2.6 in alcohol-related liver disease. In chronic viral hepatitis it is below 1.0 in most people, but it climbs as cirrhosis develops, from about 0.59 without cirrhosis to 1.02 with it. So below 1 is typical of fatty liver or viral hepatitis before cirrhosis, and once cirrhosis is present the ratio drifts above 1 whatever started it.

As a discriminator it is weak. The ranges in the founding study overlapped heavily, from 0.3 to 2.8 for fatty liver and 1.1 to 11.2 for alcohol-related disease, and no figures for how often it is right or wrong were ever published.

And the current UK guideline does not use the alcohol cut-off. It uses a ratio above 1 as an indicator of advanced fibrosis or cirrhosis, of any cause, and notes that the ratio keeps that usefulness even when both enzymes are sitting inside their reference intervals. That is a different claim from the one the rule of thumb makes, and it is the one with a guideline behind it.

It was never a test of whether somebody drinks either. The 1979 comparison it came from set people who already had alcoholic hepatitis and cirrhosis against people with other liver diseases, so a ratio above 1 in someone who does not drink points at fibrosis and not at undisclosed drinking. In fatty liver it climbs with the stage of scarring, from about 0.7 with none, to 0.9 with mild fibrosis, to 1.4 at cirrhosis.

Venous samples only

AST is accredited for venous samples and is not reported on a finger-prick liver profile. That is why our Liver Profile is 8 markers on a venous sample and 7 on a finger prick.

ALP

Alkaline phosphatase

An enzyme found in the bile ducts, in bone, in the placenta, and also in the intestine, the kidney and white blood cells. It rises when bile flow is obstructed, and separately when bone turnover is high.

Structurally it is 3 genetically distinct forms. There is an intestinal one, a placental one, and a tissue-nonspecific one shared by liver, bone and kidney. That matters for a practical reason further down, because the intestinal form is what responds to a meal.

Because of that spread of sources, an isolated raised ALP does not point to the liver on its own, and it is read alongside GGT. The UK guideline puts the logic directly. GGT is abundant in the liver and also present in the kidney, intestine, prostate and pancreas, but not in bone, so it is useful in confirming that a raised ALP is of liver and not of bony origin.

Raised levels can reflect obstructed bile flow, or bone conditions such as Paget's disease and vitamin D deficiency.

Three things raise it without anything being wrong

Adolescence. Normal bone growth in children and adolescents raises ALP to between 2 and 5 times the level seen in adults, and that is why a teenager's result has to be read against a teenage range. What bone turnover is doing, and why it is so high in adolescence, is in Minerals and bone, in the Body Guide.

Pregnancy. Placental ALP produces a 2 to 3-fold rise in the second and third trimesters of a normal pregnancy.

A meal, and your blood group. The intestinal form of the enzyme can raise serum ALP after eating, and the effect is concentrated in people with blood group B or O who are also secretors of the ABH blood group substances. That is a genuine and documented dependence and not a curiosity. What is not documented is how big the rise is. The mechanism and the blood-group dependence are established in the literature and the magnitude is not, so there is no number to give. A fasted sample removes the question.

A persistently low ALP is rarely commented on and is not nothing

A low ALP is seen in zinc deficiency, since ALP is a zinc-containing enzyme, and in magnesium deficiency. It is seen in malnutrition, and specifically in acute calorie restriction. All 3 of those are established.

An underactive thyroid appears on almost every textbook list of causes. It is not supported by primary research, and it should not be presented alongside the others as though it were equally established.

The cause that matters is hypophosphatasia, an inherited disorder of mineralisation caused by variants in the ALPL gene, and it is frequently missed precisely because a low result looks reassuring.

It is much more common than "rare" suggests

The severe form affects about 1 in 300,000 people in France and northern Europe. That is the figure usually quoted. The milder dominantly inherited form is about 1 in 6,370, roughly 47 times more common.

Undiagnosed carriage is more common still. About 7 per 1,000 people carry a known disease-causing ALPL variant, and where that was counted not one of them had a clinical diagnosis.

What the follow-up actually is

The tests are serum pyridoxal-5'-phosphate, a form of vitamin B6, and urinary phosphoethanolamine, followed by sequencing of the ALPL gene. In people with a persistent low ALP, the best thresholds are an ALP below 25 units per litre, a pyridoxal-5'-phosphate above 180 nanomoles per litre and a phosphoethanolamine above 30 micromoles per gram of creatinine. An ALP below 25 on its own separates people poorly. Adding either of the other 2 improves it substantially, and adding both makes it close to definitive.

The 2026 study that changes the practical advice

Work published in 2026 followed adults referred for bone or endocrine assessment who had a low ALP, most of whom turned out to carry a disease-causing variant. Four findings from it matter.

Every person with an ALP at or below 25 units per litre had a disease-causing variant. A higher ALP did not rule the condition out. Pyridoxal-5'-phosphate separated the groups, while bone-specific ALP, standard biochemistry and bone density T-scores did not. And each 7.8 units per litre fall in ALP was associated with a higher fracture probability independently of genotype, meaning the residual enzyme level itself carries information and not just the presence of a variant.

The reason this is more than academic is the treatment. The drugs used for osteoporosis that work by suppressing bone turnover are contraindicated in hypophosphatasia, and people have been given them before the diagnosis was made. So with a low ALP the failure mode is not a missed opportunity but a wrong treatment.

An ALP that falls suddenly is a different signal

Everything above is about a persistently low ALP. A sharp fall is not the same thing. The same large analysis identified acute low ALP alongside major trauma or surgery, multi-organ failure, massive transfusion and apheresis, and 15% of those people died within 35 days of the lowest value. That is a hospital picture and not something that turns up on a routine panel, but it is the reason a falling ALP is not read the same way as a stable low one.

Where the numbers come from

There is no published healthy threshold for this marker. The population reference range is assay-dependent, and it is more age-dependent than most of this panel, which is why a result in an adolescent has to be read against an adolescent range and not an adult one.

The one number in the literature that functions as a threshold is the 25 units per litre above, and it is a threshold for investigating a low result and not a target.

GGT

Gamma-glutamyl transferase

An enzyme concentrated in the bile ducts, and also present in the kidney, intestine, prostate and pancreas. Its established clinical use is to confirm whether a raised ALP is coming from liver or bone, since GGT rises with the first and not the second.

What actually raises it, in order

The usual claim is alcohol. The UK guideline puts obesity first. Its wording is that GGT is most commonly raised as a result of obesity, excess alcohol consumption, or drug induction, in that order. In a general UK population being screened rather than investigated, a raised GGT is more likely to be metabolic than alcoholic.

It is also true that GGT is the most alcohol-sensitive of the routine liver enzymes, which is a narrower claim than the one usually made and needs separating from it. Against the markers built specifically to detect drinking, GGT is third. Compared head to head against sober controls, phosphatidylethanol separates drinkers from non-drinkers best, carbohydrate-deficient transferrin next, then GGT, with mean cell volume last and barely behind it.

For the light to moderate drinking that most people taking a blood test are actually asking about, GGT performs poorly. The evidence behind that is strongest in pregnancy, where no blood marker picked up reported drinking reliably, and it has not been tested as directly in the general adult population. It either misses most drinkers or flags people who are not drinking, depending on where the line is drawn. A normal GGT does not exclude regular drinking.

A raised GGT in a person who does not drink

Specificity for alcohol has been put as low as 40%, so in a population where most people do not drink, most raised results have nothing to do with alcohol, and the guideline's own ordering puts obesity ahead of it. That does not make the result meaningless. Of the routine liver enzymes GGT is the strongest predictor of death from liver disease, and the only one that predicts death from any cause. A raised GGT in a non-drinker is a metabolic and liver finding rather than a question about honesty.

Several common medications raise it, which is the third item on the guideline's list and the one most often forgotten when a result comes back unexpectedly.

Where the numbers come from

No professional body publishes a healthy threshold or a target for GGT. The population reference range is assay-dependent.

Beyond the established use above there is a large body of observational work that needs describing properly, because it is quoted more confidently than it deserves in both directions.

The cardiometabolic association is real and the numbers are large

A higher GGT carries roughly two-thirds more cardiovascular death in both men and women. Under the age of 60 that doubles in men and more than doubles in women. The association holds for chronic coronary disease, heart failure and both kinds of stroke, and not for acute heart attack in men.

It climbs steadily with the number, at about 10% more cardiovascular death for every 10 units per litre, and it does so inside the reference interval. So the gradient runs through the normal range and does not start at the upper limit.

The direction of travel matters as well as the level. A GGT rising by more than 9.2 units per litre over 7 years carries about 40% more cardiovascular death in men than a stable one, and that holds inside the normal range. In UK adults the gradient runs at roughly a quarter more risk without diabetes and 40% more with it.

And it is almost certainly not causal

That same UK analysis found that knowing someone's GGT does not improve cardiovascular risk prediction beyond the traditional risk factors. So the association is reproducible and adds nothing on top of what a standard risk score already knows.

The genetic evidence cuts against GGT being the thing doing the damage. Inherited variants that raise a marker for life make it possible to test whether the marker itself causes the harm. Variants that raise GGT carry no measurable increase in coronary disease, heart attack or diabetes. In the same work, variants that raise ALT were strongly linked to diabetes, so the method was capable of finding an effect.

A larger analysis testing the question in both directions found no evidence that GGT causes type 2 diabetes, and supported a sequence running from insulin resistance to liver fat to diabetes, with ALT and AST as the markers that track it.

Lower is not reliably better either

The reasonable summary used to be that lower is generally better within the range. Newer work complicates it. The ratio of GGT to HDL cholesterol has a J-shaped relationship with death. Below a threshold there is no association at all, and above it each near tripling of the ratio carries about a third more death from any cause and two-thirds more cardiovascular death. A J shape is an argument against treating lower as automatically better.

No intervention study exists

The straight answer to the obvious question is that no trial has ever tested whether lowering a within-range GGT changes any outcome. Not one. What a rising GGT is genuinely good for is prompting a look at weight, at alcohol intake and at any new medication, which are all things that repay a look on their own account.

Recent work, and what it does and does not add

The 2025 and 2026 literature on GGT is prognostic and mechanistic, and none of it is interventional.

The mechanism proposed for all of this is GGT's job of breaking down glutathione outside the cell, which generates reactive oxygen species when transition metals are around. In transthyretin cardiac amyloidosis, GGT is by far the most deranged liver marker, raised in 48% of people against 1% for ALP, and it predicts death and heart failure admission beyond the established risk score. It also statistically accounts for 5% to 12% of the link between markers of insulin resistance and later cardiovascular disease in adults with depression, but this is not the same as showing it causes any of it and has not been shown in a general population. Insulin resistance itself is in Blood sugar and metabolism, in the Body Guide.

Taken together, the fair reading is that GGT is a marker of underlying metabolic and oxidative stress and not a target in itself.

Total bilirubin

The pigment produced when red blood cells are broken down. The liver processes it and excretes it in bile, so it rises either when red cells are being destroyed faster than usual or when the liver is not clearing it efficiently. How red cells are made and broken down is in Blood cells, in the Body Guide.

Fasting raises it, in everyone, and roughly doubles it within 24 hours in people with Gilbert's syndrome, a common and harmless inherited variation in how bilirubin is processed. It is also reported to rise with illness and with stress.

Gilbert's syndrome is far more common than usually stated

The figure usually quoted is a small one. That describes the diagnosed minority and not the condition.

About 10% of people carry the genotype that defines Gilbert's, and only 3% of those carriers have a recorded diagnosis. The visible form, where the bilirubin is actually raised, runs at about 8%. So the genotype is around 1 in 10 people of European background, the clinically apparent version is lower, and the recognised diagnosis is rarer still.

The mechanism, and how much of bilirubin it explains

The gene is UGT1A1, and the variant is an expansion in its promoter region, the stretch of DNA that controls how much of the enzyme gets made. The effect is to reduce the liver's capacity to attach a sugar group to bilirubin, which is the step that makes it water-soluble and excretable. All 18 well-characterised people with Gilbert's in one study carried it.

Variation at that one location explains 37.1% of all the variation in measured bilirubin across a population.

The fasting effect, and its 2 catches

The original work, from 1973, restricted people to 400 calories a day for 24 to 72 hours. A rise of 100% or more, so a doubling, was taken to indicate that unconjugated bilirubin was raised because of Gilbert's, and the rise was usually visible within 24 hours of cutting the calories. In the Egyptian cohort, fasting for more than 12 hours, notably during Ramadan, brought on visible jaundice in 72% of people with the condition.

The first catch is that fasting raised bilirubin in the control group too. It is the size of the rise that separates the groups and not the direction, which is a much weaker test than it sounds. The second is specificity. Both the fasting test and the alternative rifampicin test raised unconjugated bilirubin in nearly half of a group of people who had chronic liver disease instead. The 1973 work that introduced the fasting test found the opposite, reporting no significant rise in people with liver disease, so the 2 primary sources disagree on this point.

The fasting test is no longer how the diagnosis is made. Genotyping is.

Where the numbers come from

There is no published healthy threshold for this marker. The population reference range is assay-dependent.

The practical point follows from the fasting note above, and it needs spelling out because our own instructions interact with it. We ask for several of these panels to be taken fasted. Fasting is one of the things that raises bilirubin in someone with Gilbert's syndrome.

So a mildly raised bilirubin on a fasted sample, with everything else in the liver profile normal, is a recognised and benign pattern and not a puzzle. It is better to know that before the report arrives than after. The UK guideline is unambiguous on the substance. Where the excess is predominantly unconjugated and red cell breakdown has been excluded, the cause is virtually always Gilbert's syndrome, and such people should be fully reassured.

One measurement point. Haemolysis in the sample can push bilirubin the wrong way, and usually downward, because haemoglobin interferes with the colour reaction the assay relies on. On at least one platform the bias runs upward instead. A haemolysed sample is not a bilirubin result.

The idea that a high-normal bilirubin protects your heart, and why it does not survive

This is one of the most widely repeated claims in the optimisation space, and the evidence on it has moved decisively in the last few years.

The observational case is real. A higher bilirubin consistently tracks with less cardiovascular disease, and it has been argued that people with Gilbert's are protected from a range of conditions of affluence, including cardiovascular disease, some cancers, and autoimmune and neurodegenerative conditions.

The genetic test breaks it, and it breaks it in 2 separate populations. Higher measured bilirubin is strongly associated with poorer overall health, with chronic obstructive pulmonary disease, with heart attack and with cholesterol, exactly as the observational work says. Those associations are absent in the people who carry the Gilbert genotype. So bilirubin has no likely causal role in protecting against any of them, and it is a poor thing to try to change.

The reason the observational signal exists appears to be that the causation runs the other way. Bilirubin is consumed by oxidative stress, so illness lowers it, and the sicker the population the lower the average bilirubin. There is direct evidence for that. In multiple sclerosis, bilirubin runs lower than in the general population, at 8.3 against 9.6 micromoles per litre, while the UGT1A1 genotype frequencies in the 2 groups are identical. The difference could not have been genetic. Bilirubin also runs lower in type 2 diabetes, but the picture there is not identical, because the UGT1A1 genotype frequency differs in men, so genetics contributes something alongside consumption under oxidative stress.

So the association is solid and the causal claim is refuted. A high-normal bilirubin is not a cardiovascular asset, and raising bilirubin is not a goal. One thing the genotype does appear to carry is a modestly higher chance of gallstones, at roughly 16% more.

Total protein

All the protein circulating in blood, essentially albumin plus the globulins. It is a broad indicator of nutritional state, liver synthetic function and hydration.

Where the numbers come from

There is no published healthy threshold for this marker. The population reference range is assay-dependent.

Read alone it says very little, and the reason is arithmetic and not a study finding, so it should be put that way. Total protein is albumin plus globulin. A fall in albumin offset by a rise in globulin leaves the total unchanged, and that combination is exactly what chronic inflammation, chronic liver disease and a paraprotein disorder such as myeloma produce. The total can therefore sit in the middle of its range while both of its components are abnormal.

That is the explicit reason the albumin to globulin ratio is calculated instead of the total being read directly. No study has been done to test the masking effect, because none is needed for a subtraction, and the honest description is reasoning and not evidence.

It is not useless. It is the input that makes the ratio possible, and a high globulin fraction is what raises the possibility of a paraprotein in the first place. It earns its place as the other half of a pair.

Albumin

The most abundant protein in blood, made by the liver. It maintains fluid balance between blood and tissue, and carries hormones, calcium and many drugs around the body.

Because it carries calcium, it is required to interpret a calcium result correctly, which is what adjusted calcium does.

It falls with liver disease, malnutrition, inflammation, sepsis, loss of protein through the gut, and kidney protein loss, where heavy protein loss into the urine is called nephrotic syndrome. It rises with dehydration.

The UK guideline lists the situations in which albumin is reduced as including sepsis, systemic inflammatory disorders, nephrotic syndrome, malabsorption and gastrointestinal protein loss, and it warns specifically against overinterpreting albumin as a marker of how severe liver disease is. That warning is the frame for this whole entry.

The rise with dehydration is not a rise at all. Nothing is being made faster. The same amount of albumin is sitting in less water, so the concentration reads higher. That is the same artefact posture and a tight tourniquet produce, which is the next point.

Posture moves it, by more than you would expect

Standing up shifts water out of the circulation and concentrates everything left behind. Plasma volume falls by about 3.4% going from lying to sitting, by about 14.1% going from lying to standing, and by about 9.7% going from sitting to standing.

For albumin specifically, the measured change from sitting to lying is a 6.5% fall, which is a clean posture measurement. The 6.3% rise usually quoted for lying to sitting was taken after an hour of moderate activity, so it combines posture with exercise. The rise depends on molecular size, which means it is not purely a water shift, while the fall on lying down does not.

The time course is the part that is practically useful. Sitting for 30 minutes rather than lying raises albumin by roughly 5 to 7%. Thirty minutes of slow walking adds a further 10 to 14%. Returning to a sitting baseline takes about 20 minutes of quiet sitting, which is where the general recommendation to sit for at least 20 minutes before a blood draw comes from.

It is among the worst affected and it is not alone

Albumin is often described as uniquely dilution-sensitive. In the same experiment that produced the figures above, haemoglobin, haematocrit, total protein, ALP, amylase, AST, total bilirubin, calcium, total and HDL cholesterol, GGT, glucose, LDH, magnesium and triglycerides all showed meaningful bias between sitting and standing. Albumin exceeded the quality specification for bias at every posture change, so it is among the worst affected. It is not in a category of its own, and the 20-minute rule applies to the whole panel and not only to this line.

Where the numbers come from

No professional body publishes a healthy threshold for albumin. Searching for one returns consumer testing and supplement sites and no medical society anywhere. The population reference range is assay-dependent, and the UK guideline warns against overreading it.

Two things about how the number behaves. Because it falls with inflammation, a low albumin is frequently not about the liver at all. And because of the posture effect above, small differences between 2 samples from the same person can have nothing to do with the person.

The within-range outcome evidence, and its ceiling

Low-normal albumin does predict mortality, and the effect appears at levels inside the reference range.

Albumin values sitting between the 10th and 90th percentiles of the reference interval are associated with death from any cause, after allowing for age, sex, ethnicity, income, chronic illness and prior healthcare use. ALP behaves the same way.

A higher albumin inside the clinical normal range tracks with less death from any cause, by roughly a third in men for each 2.6 grams per litre and a fifth in women for each 2.4. That only held in people whose cholesterol was at or above the median, with no association at all below it. An albumin at or below 43 grams per litre came with a much higher cardiovascular death rate in men, on figures too imprecise to act on.

The ceiling on all of this is the figure from the ALT entry. The same work that found within-range albumin associated with death found that all 152 markers it tested, put together, explained an additional 0.8% of the variation in that risk. A low-normal albumin is a marker of inflammation and frailty, it is reproducibly associated with worse outcomes at population scale, and no study has shown that raising it changes anything.

Albumin is also one of the 3 inputs to calculated free testosterone, alongside total testosterone and SHBG.

Globulin

Calculated as total protein minus albumin. It represents the remaining protein fraction, which includes antibodies and various transport and inflammatory proteins.

The albumin to globulin ratio is more informative than either number alone, and unusually for a calculated line on a report, there is real outcome evidence behind it.

Where the numbers come from

There is no published healthy threshold for this marker, and no professional body publishes a target for the ratio either. The population reference range is assay-dependent.

As with the unsaturated iron binding capacity in the iron group, this is a calculated figure and not a measured one. It is a reported line, and it earns its place as a pair with albumin and not as an independent measurement.

What the ratio predicts

The best evidence in a general population comes from generally healthy Korean adults aged 30 and over, followed for about 6 years. These are the figures, compared against a ratio of 1.5 or above.

A ratio between 1.0 and 1.1 came with roughly 2.7 times the death rate from any cause, 3 times the cancer death rate and twice the chance of developing cancer. Below 1.0 those rose to about 6.7, 4.4 and 4 times.

The obvious objection is that it is picking up cancer that was already there and not yet diagnosed. That was tested. Excluding every cancer diagnosed in the first 2 years still left roughly twice the risk. The associations were strongest for liver and blood cancers.

Three things that keep it from being a test

Those findings have never been replicated at a larger scale, and nobody has established how often a low ratio is right or wrong. There is no published figure for that at any threshold, so the ratio has no validated cut-off to read a result against.

A near 7-fold figure, imprecisely measured, in people who referred themselves for health screening, invites confounding by chronic disease that was present and undiagnosed.

And almost the entire recent literature on the ratio, including nearly all of the 2025 and 2026 work, is about predicting outcomes in people already diagnosed with cancer, in colorectal, oesophageal and gastric disease. That does not transfer to screening people who feel well.

The thresholds of 1.5, 1.1 and 1.0 come from that primary research and from nowhere else. The reasonable reading is that the ratio is a non-specific marker of chronic inflammation with genuine outcome data behind it, and not a test.

Lipase

An enzyme produced by the pancreas to digest fat. It is the preferred marker for pancreatic inflammation, having largely replaced amylase. What the pancreas does, and how its 2 jobs differ, is in Liver, in the Body Guide, under Pancreas.

It is not a liver marker, but it sits naturally alongside them because upper abdominal symptoms often need both.

Why it replaced amylase, and what it did not win on

The usual explanation is that lipase is both more specific than amylase and stays raised longer. The second half is right and is the real reason. The first half is not.

At the standard threshold of 3 times the upper limit of normal, in people arriving at hospital with abdominal pain, lipase picks up about 79% of pancreatitis and amylase about 72%. A urine test for a third enzyme, trypsinogen-2, matches amylase.

So lipase catches more cases. On how often each is wrong in people who turn out not to have pancreatitis, the 2 are too close and the data too thin to separate them. What cannot be said is that lipase is the more specific test.

What lipase does win on, besides catching more cases, is duration. Its longer elevation gives a wider diagnostic window, and that advantage is largest in pancreatitis caused by alcohol. Amylase has 2 problems lipase does not. Its window is short, and it throws false positives from salivary amylase and from a harmless bound form called macroamylasaemia.

Two figures matter here. About a quarter of genuine cases are missed at the standard threshold, and about 1 in 10 people without pancreatitis are wrongly flagged. So the threshold for admitting and treating someone should be low where the symptoms are suggestive, even when these tests come back normal. Accuracy also falls the longer it has been since the pain started.

A raised lipase with a normal amylase usually is not pancreatitis

Where a lipase above 3 times the upper limit sits alongside a normal amylase, the explanation is usually something else entirely. It is reduced kidney function, a tumour outside the pancreas, gallbladder inflammation, oesophagitis, a delay in getting the blood to the lab, or a high triglyceride level. A raised lipase with a normal amylase is not evidence of pancreatitis.

Where the numbers come from

The committed number for lipase is a diagnostic threshold and not a health one. It is 3 times the upper limit of the lab's own reference range.

The criteria it belongs to need stating exactly, because the usual paraphrase is wrong in a way that matters. The revised classification of acute pancreatitis, agreed by international consensus in 2012 and published the following year, requires any 2 of 3 things. It came out of a web-based consensus across 11 national and international pancreatic associations rather than from any single professional body. The first is abdominal pain consistent with acute pancreatitis, meaning sudden and persistent severe pain in the upper abdomen often radiating to the back. The second is a lipase or amylase at least 3 times the upper limit of normal. The third is characteristic findings on a contrast-enhanced CT scan, or less commonly on MRI or ultrasound.

Any 2 of 3, and not enzymes alongside symptoms. Enzymes plus imaging is sufficient without typical pain. Enzymes on their own never are.

Two further things about that threshold need to be clear. It is defined against a lab range and not as an absolute figure, so it moves with whichever range your report carries. And 3 times is a consensus point inside a range and not a physical boundary. At least one published source recommends 2 to 4 times the upper limit.

There is no published figure for what a good lipase looks like in a healthy person, and the population reference range is assay-dependent.

A persistently raised lipase in someone who feels well

There is a named condition for this. Chronic asymptomatic pancreatic hyperenzymaemia, sometimes called Gullo's syndrome, is a long-running elevation of amylase, pancreatic isoamylase, lipase and trypsin with no pancreatic disease behind it. It can run in families, it persists over time, and it fluctuates considerably, including frequent returns to normal.

It is usually benign, and the imaging literature argues against dismissing it without a look. Where people with asymptomatic raised enzymes had MR imaging of the pancreatic ducts, the pancreas was abnormal in 57%, against a much lower rate in controls. There was an anatomical variant called pancreas divisum in 18.5%, changes of chronic pancreatitis in 16.6%, and small cysts in 27.8%. No cancers were found. Using endoscopic ultrasound instead, abnormalities appeared in 60.3% against 13.2% of matched controls. Imaging is the recommendation before calling the finding benign.

Whether this marker belongs on a panel for people who feel well

Said openly, the case for measuring lipase in someone without symptoms is weak, and the reasons are specific.

There is no healthy or optimal figure published by anyone. There is no outcome evidence for lipase in people without symptoms. The diagnostic criterion it is measured against presupposes someone presenting with abdominal pain. And around 60% of people with an incidental elevation turn out to have some finding on detailed imaging, which then has to be worked up.

It is on the panel because upper abdominal symptoms are common, because the marker is cheap, and because a genuinely high lipase in someone with pain needs acting on the same day. Those are reasonable grounds. They are not the same as evidence that measuring it in someone who feels well does any good.

One note on the alternatives, since it comes up. Urinary amylase does not outperform the blood test, and its claimed usefulness in people who present late is unsubstantiated.