Vitamin and Mineral Markers

Vitamin D

25-hydroxyvitamin D, 25-OH D

The storage form of vitamin D, and the standard way vitamin D status is assessed. It reflects both what the skin makes from sunlight and what comes from diet and supplements. Why the storage form is measured instead of the active hormone, what the receptor does, and the unsettled question of how much of the total is actually available to tissue are all in Nutrients, in the Body Guide.

Severe deficiency causes rickets in children and osteomalacia in adults, and long-standing deficiency contributes to osteoporosis.

The trend in rickets gets stated wrongly in both directions. UK hospital admissions for it climbed from the late 1990s to about 2011 and have been broadly flat since, at several hundred a year in England. It is not rising now, and it has not gone away. The cases are not spread evenly either, and most are in Black and South Asian children.

Osteomalacia is the part that matters for reading this panel, because it is widely believed to show as a low adjusted calcium. It usually does not. Parathyroid hormone rises early and holds calcium in range, so a low calcium is a late feature and often never appears at all. Measured in UK adolescents where most were below 25, the median adjusted calcium was normal in every group and alkaline phosphatase was raised in about 15%. A normal calcium and a normal alkaline phosphatase alongside a low vitamin D say nothing about bone.

In the UK levels fall through autumn and winter for almost everyone, because from about October to early March the sun is not strong enough for the skin to make any meaningful amount. Time of year matters more than time of day, and it matters more than most of the things people worry about instead.

Where the numbers come from

The Scientific Advisory Committee on Nutrition, whose position NICE adopts, sets 25 nanomoles per litre as what it calls a population protective level. Its own wording is that this is the concentration individuals in the UK should be above, throughout the year, in terms of protecting musculoskeletal health, and that it is not a clinical threshold diagnostic of disease. It recommends 10 micrograms, which is 400 international units, daily for everyone in the UK aged 4 and over.

One detail about that recommendation gets reversed constantly. The committee says the intake applies throughout the year, as a precautionary measure. The familiar advice to supplement particularly in autumn and winter is government implementation advice on top of it, and not the committee's own position.

That figure is a floor, not an optimum, and the distinction matters because the two get confused constantly. It is not a target, and the committee sets nothing above it.

The band many labs print between 25 and 50 does have a UK body behind it. The Royal Osteoporosis Society proposes a 3-band scheme for clinicians in respect of bone health, putting below 25 as deficient, 25 to 50 as possibly inadequate in some people, and above 50 as sufficient for almost the whole population. It describes those thresholds, in its own word, as pragmatic, arrived at after reviewing the American and UK reports. The same document says universal screening of asymptomatic populations is not recommended, which is a position a company selling the test should print. Where the argument about the right level currently stands, including a UK position statement arguing the floor should move to 50 and the American body that withdrew its own 75 figure in 2024, is set out in Nutrients, in the Body Guide. The rest of what follows is about the measurement, which is a separate problem and a larger one than most people assume.

The same tube of blood, 445 labs, and a spread of a quarter

Vitamin D has the best international quality scheme of any marker on this menu. It is run from a London hospital, it distributes the same serum to over 1,000 participants every quarter, and the reference value for each sample is established by the American reference method. In the October 2024 distribution, one ordinary serum sample with a reference value of 46.7 came back from 445 labs with a mean of 45.6. The mean was almost exactly right. The scatter around it was 13.5%, which puts the great majority of those labs somewhere between roughly 34 and 58 for the same blood.

That is a quarter of the way up and down from the true value, on a marker where the UK decision line is 25 and the lab convention band runs from 25 to 50. Nothing was wrong with the sample and nothing was wrong with the reference method. The labs simply do not agree with each other, and the one that ran yours is the one whose answer you got.

What standardisation fixed, and what it did not

There has been a serious international effort to standardise vitamin D measurement, and it worked, for a particular definition of worked. An assay counts as standardised when its mean bias against the reference method is under 5% and its mean coefficient of variation is under 10%. Getting the average right is not the same as getting your sample right, and once the averages were fixed the scatter on individual samples became visible.

The American certification programme therefore started publishing how often each method lands within 5% of the reference value on individual blinded samples. Across 34 certified methods, mass spectrometry managed a mean of 63%. The immunoassays, which is what almost every routine result is, managed a mean of 30%, with the range running from 8% to 68%. So a fully certified immunoassay is typically outside 5% of the true value on 7 samples in 10.

And mass spectrometry is not automatically the right answer either. In an international comparison of 14 mass spectrometry methods as they are actually built and run in real labs, only about half met the same 5% average bias criterion, and 4 of them were 12% to 21% out. The reference methods are excellent. The in-house methods built to imitate them vary a great deal. "Measured by mass spec" is a reason for mild optimism and not a guarantee.

The molecule that looks identical and is not

There is a mirror-image version of 25-hydroxyvitamin D circulating in blood, called the C3-epimer. It has the same mass as the real thing, so mass spectrometry cannot tell them apart by weight and has to separate them physically before measuring. In most adults it is low, often below the level an assay can quantify at all, and a very small fraction of the total. In infants it is a great deal higher.

Two findings from the October 2024 distribution matter here. Of the 90 mass spectrometry labs taking part, 44% separated the epimer properly and 56% did not, and the ones that did not over-report as a result. And among the automated immunoassays, most do not see the epimer at all, with 2 exceptions. One of them is the third-generation assay from the largest manufacturer, which the scheme measured as reading the epimer at 91% of the strength of the real thing, close enough to count it as if it were vitamin D. The manufacturer's own method sheet puts its cross-reactivity higher still, at 122%, measured a different way. Current advice is to exclude the epimer from a vitamin D result. Whether your result excluded it depends entirely on which method ran it.

The scheme ran the experiment that shows what this costs. Alongside the ordinary serum described above, it distributed the same serum with 50 nanomoles per litre of the epimer added. The labs came back with a mean of 62.1 and a scatter of 43.9%, against 13.5% on the sample without it. Adding 1 molecule that half the methods cannot see more than tripled the disagreement between labs.

Pregnancy makes immunoassays disagree, and only in pregnancy

Almost all of the vitamin D in blood is locked to a carrier protein, and every immunoassay has to prise it off that carrier before the test can see it. How completely it manages that depends on how much carrier protein is there. Pregnancy raises the carrier substantially, so immunoassays tend to under-read in pregnancy, and different assays under-read by different amounts.

The demonstration of this is neat. Three automated immunoassays run on the same third-trimester samples gave results that differed significantly from each other, with something like 24 nanomoles per litre between the highest-reading and the lowest-reading assay. The same 3 assays, run on non-pregnant women, showed no significant difference at all. The assays are not unreliable in general. They come apart in pregnancy, and they come apart in the direction of reporting deficiency that is not there.

Biotin. Vitamin D is on the list of markers biotin affects, and Biotin at the top of this guide has the detail, but this one deserves a correction more than a warning. The manufacturer of the most widely used platform tested its vitamin D assay to 600 nanograms per millilitre of biotin and found no interference, which is more than 10 times the limit for most of the other affected tests, and its method sheet carries no instruction to wait a set time after a dose. A 20 milligram daily biotin habit peaks at around 355 nanograms per millilitre in the hour after a capsule, which is still below what the assay was tested to. Above that tested limit the sheet says only that the result will read high. So the general biotin warning is well founded for the thyroid and hormone tests and it is a much weaker concern for this one.

Two interferences that matter because they are so specific

Abnormal antibody proteins in blood, the kind produced in myeloma and in the far more common and usually harmless condition that precedes it, can produce wildly high vitamin D results on some platforms. In one series 7 people were found in 4 months with results above 400 nanomoles per litre, and when 50 people with known abnormal proteins were tested, 8 read falsely high on one manufacturer's assay, 1 on another and none on a third. The same work specifically looked for the generic immunoassay culprit, the heterophile antibody, and ruled it out. If you see heterophile antibodies blamed for wrong vitamin D results, that has been checked and it is not the mechanism here.

The second is the sort of thing that explains why a number is a measurement and not a fact about your body. Fluorescein is the dye used in eye angiography. One widely used vitamin D assay is built on beads coated with an anti-fluorescein antibody, and the dye pulls that assembly apart. A patient measured 352 nanomoles per litre shortly after an eye appointment, against 34 a month earlier and 45 a further 3 months on. The dye clears in 2 to 3 days. Three people were affected before anyone worked out why.

The month of the test is one of the 2 biggest things determining the answer

Measured properly, by taking blood from the same 125 adults in Manchester every month for a year, the UK seasonal cycle runs from a mean of about 71 nanomoles per litre in September to about 46 in February. That is a swing of roughly 25, which is the full width of the lab convention band.

What that does to classification has been quantified in a UK cohort of nearly 450,000 adults. Results below 25 ran at 4.5% of summer samples and 23.1% of winter ones, so roughly 5 times as many winter samples were classified deficient. The effect survived adjustment for sex, age, body weight, ethnicity, smoking, alcohol, supplements, deprivation and region, and on that adjusted comparison it was larger than obesity, larger than smoking and larger than living in the most deprived fifth of the country. Only 1 factor in the whole analysis was bigger, and it was Asian ethnicity. Being in Scotland rather than South West England roughly doubled it too.

So a repeat test 3 months later is not a repeat test

The top of this guide explains the reference change value, the amount 2 results have to differ by before the difference means something. For vitamin D that framework does not apply, and the body that would normally publish the figure has said so. The calculation assumes a marker wanders randomly around a personal set point. Vitamin D does not. It has a systematic annual trend running underneath it, and when a careful European study sampled 91 healthy people weekly for 10 weeks it found they were not in a steady state at all, and could only be made to look like one by discarding more than half the data.

The measured drift is large. Through a European spring it ran at about 2.8% a week, adding up to 31.6% over 10 weeks. Across an autumn in Belgium, a group taking nothing fell by 39.6%. So a change of 30% or 40% between 2 results 3 months apart is the expected seasonal path and not a signal, and the same committee calculated that a common routine assay is not precise enough to separate a supplement effect from the seasonal drift anyway.

If you are going to retest, the interval matters more than the result

25-hydroxyvitamin D has a half-life of about 2 to 3 weeks, which is why it is the marker of supply and the active hormone, with a half-life of under 4 hours, is not. Starting or changing an intake produces a measurable change within a few weeks and settles at a new level over roughly 3 months. Both the UK osteoporosis guideline and the international lab medicine committee put the minimum sensible retest interval at 3 months, and the osteoporosis guideline adds that it may be more prudent to wait until 6 months have passed. Retesting sooner tells you very little, and retesting in a different season tells you about the season.

Ethnicity, and a claim that turned out to be an assay artefact

We are not aware of any body that publishes a different vitamin D threshold by ethnicity. The Scientific Advisory Committee on Nutrition lists people with dark skin and people who cover most of their skin among those at increased risk, and then states that a separate intake recommendation is not required for those groups. In practice the numbers come out very differently. In that UK cohort of 450,000, results below 25 ran at 12.0% in white participants, 34.9% in black participants and 53.7% in Asian participants. Narrow it to South Asian adults specifically and it is worse again. In the largest analysis of that group, 55% were below 25 and 92% were below 50, so more than half sit under the UK deficiency threshold and almost nobody clears the higher one.

That Asian figure hides a reversal. Split into the 3 largest South Asian groups in the same cohort, results below 25 ran at 43.5% in Bangladeshi participants, 52.0% in Indian participants and 65.7% in Pakistani participants. Bangladeshi people came out best of the 3 rather than worst, which is the reverse of the usual assumption, and it held after adjustment for age, sex, body weight, season and supplement use. Reported oily fish intake is far higher in the Bangladeshi group, which is the likeliest explanation and has not been shown to be the cause. The gap between the groups is only about 7 nanomoles per litre, so all 3 are deficient by any UK definition, and better here means less badly off rather than sufficient.

There was a well-known argument that this did not matter, and it collapsed. A 2013 study found the vitamin D carrier protein was about half as abundant in black Americans as in white Americans, which would mean similar amounts of usable vitamin D despite very different totals. It was influential, and it appeared to resolve a real puzzle, because black Americans have lower total vitamin D alongside higher bone density and fewer fractures. Three years later 2 independent groups using mass spectrometry, a different antibody and a protein-profiling method all found no difference in carrier protein by ethnicity. The original antibody had been failing to detect one common inherited form of the protein, so it was measuring genotype and not concentration. The free fraction is genuinely lower in black Americans, not equivalent, and the bone puzzle is not yet fully explained. Anyone still saying that people with dark skin need less vitamin D because their carrier protein is lower is repeating a measurement error from over a decade ago.

The 6 times figure for how much longer darker skin needs in the sun does not come from anywhere solid either. Its origin is a letter published in 1982 describing 5 people, in which a single dark-skinned volunteer was re-exposed at 6 times the original dose and came out level with the lighter-skinned volunteers. That is 1 person, measured once. The direction is not in doubt, but the size of it is, and how much sun a dark-skinned person actually needs to reach sufficiency has never been established.

It gets less tidy still. Controlled ultraviolet experiments have found the rise after an identical dose predicted by the starting vitamin D level rather than by measured skin pigmentation, and matching a darker-skinned group to a lighter-skinned group on starting level removed the difference between them. The defensible UK number comes from giving the same simulated British summer sunlight to South Asian and to white adults, where the rise was about 2.4 times smaller in the South Asian group. That is a long way from 6 times, and further still from 10.

There is a second consequence, and it runs the other way from what you would expect. Recent screening in northern England found insufficiency rates in older adults and in adults with darker skin that did not fall at all over the summer. If you make very little in July, you have very little to lose by January. Seasonal reasoning about a result is most useful for the people who need it least.

Clothing has been measured here rather than assumed, and what came out is narrower than the assumption. UK work recorded what people wore day by day and converted it to the share of body surface left exposed. Shorts and a t-shirt come to about 23%. A long-sleeved top, trousers or a long skirt and a head covering come to about 4%.

Applied to adolescents in northern England, South Asian girls had 8 to 10% of their skin exposed on summer weekends against 15% in white girls. South Asian boys exposed the same share as white boys, so this was a difference between the girls rather than between the ethnic groups. Time outdoors on weekdays was the same in both groups. What separated them was weekends and holidays, with far fewer of the South Asian adolescents going away at all. Sunscreen was not the mechanism, because almost nobody in either group used any.

No UK body recommends testing any of these groups, and we sell this test. NICE says health professionals should not routinely test vitamin D status unless a person has symptoms of deficiency, a clinical reason such as osteomalacia or a fall, or a particularly high risk such as very low sunlight exposure. The Royal Osteoporosis Society lists the risk factors that predispose to a low result and then says not to routinely test the people who have them. Migrant health guidance on GOV.UK says routine testing is not recommended. All 3 name people of African, African-Caribbean and South Asian family origin, and people who cover most of their skin, as groups to supplement rather than groups to measure.

What those groups are advised instead is to consider 10 micrograms daily all year, rather than only through autumn and winter like everybody else. So the guidance already treats ethnicity as a reason to act. It does not treat it as a reason to test.

Both of those can hold together. A population recommendation is a judgement about millions of people who will never see a result, and it turns on cost, on whether the answer changes what people do, and on evidence that testing improves outcomes, which for vitamin D does not exist. Wanting to know your own number is a different question, and the answer can be useful to you without being worth recommending to the country. What it will not do is change the advice, which is to supplement either way. If that makes the test uninteresting to you, that is a fair conclusion to reach, and we would rather you reached it before buying than after.

It is possible to have too much, and the real figures are further away than the warnings suggest

Excess vitamin D raises calcium, and that is the route to nausea, thirst, kidney stones and, over time, kidney damage. It comes from supplements and not from sun or food, because the skin stops producing once it has enough.

The honest version of the risk needs the numbers. The level at which toxicity is generally defined is around 375 nanomoles per litre, roughly 5 times the top of a normal British summer. In a review of the published toxicity cases the doses ran from 50,000 international units a day upward, with durations from 2 weeks to 3 years. The tolerable upper intake level used in the UK is 4,000 international units a day, which is the European figure, the Scientific Advisory Committee on Nutrition having set none of its own.

That is a wide margin and it is not a guarantee. Toxicity has also been reported at much more modest intakes in people with an inherited difference in how they clear vitamin D, and in conditions such as sarcoidosis that raise sensitivity to it. So the distance between the upper limit and the doses in the case literature describes the ordinary situation and not everybody's.

A 10-year American population series puts a number on how alarming this should be. Among 20,308 vitamin D measurements, 8.4% were above 125 nanomoles per litre and 37 people were above 250. Across the whole set, the vitamin D result was not significantly related to the calcium result or to the risk of a raised calcium. Four people had a high vitamin D that coincided with a raised calcium, and 1 had clinical toxicity, at about 910. Over the decade the number of high results rose enormously with no matching rise in toxicity.

Where those published cases come from also matters. They are dominated by prescribing errors, compounding errors, deliberate megadosing and contaminated products. One recent family of 3 presented together with severe hypercalcaemia traced to unpackaged cooking oil that turned out to be contaminated with vitamin D3. We were not able to identify a case series of toxicity from lawfully sold, correctly labelled supplements taken at the dose on the label, though prescribed high-dose preparations do appear in the recent literature. High-dose supplementation without testing is still the route people take to get there, and 400 international units is not that.

Ergocalciferol, and a warning that has expired

Vitamin D comes in 2 forms. D3 is what skin makes and what most supplements contain, and D2, ergocalciferol, is what some prescriptions and many vegan supplements contain. For years the fair warning was that immunoassays under-read D2, so people taking it got falsely low results. That was a reasonable description of the assays of 10 to 15 years ago, and it was never uniform, with the direction of the error differing between manufacturers on the same samples. Tested again in 2026 against mass spectrometry, including samples deliberately chosen from people on high-dose D2, 4 current assays had a mean overall bias of about 13 nanomoles per litre or less, and the finding was that D2 recovery no longer appears to be a significant concern for the assays evaluated. The part of the old warning that survives is that individual samples can still be well out, and that the disagreements cluster right at the classification thresholds.

Units

American reporting uses nanograms per millilitre. A result of 25 nanomoles per litre is 10 nanograms per millilitre, and 50 is 20.

Total Vitamin B12

All the B12 in the blood, both the fraction bound to a carrier that delivers it to cells and the larger fraction bound to a carrier that does not.

This is its weakness. Most circulating B12 is bound to haptocorrin and is not available to tissue, so total B12 can read normal while the usable fraction is low. Only about 20% to 25% of it is on the delivering carrier. The route B12 takes into the body, why deficiency usually fails at absorption and not at intake, and the 2 jobs B12 does inside the cell are all in Nutrients, in the Body Guide.

Deficiency causes anaemia with large red cells, and separately causes nerve damage, meaning numbness, pins and needles and balance problems, which can appear before the blood count changes and is not always reversible. Pernicious anaemia, an autoimmune condition affecting absorption, is a common cause.

Total B12 remains the standard first-line test and is adequate for most people. Active B12 is the more precise alternative in the sense that matters biologically, though not in the sense of being a steadier repeat measurement, which is covered in the next entry.

Where the numbers come from

NICE guideline NG239, published in March 2024, gives an indeterminate band instead of a single cut-off. A total B12 below 180 nanograms per litre, which is 133 picomoles per litre, is confirmed deficiency. Between 180 and 350, which is 133 to 258 picomoles per litre, is the range in which the test has not answered the question. Above 350, or 258 picomoles per litre, deficiency is unlikely.

What NICE says to do with each is not symmetrical. Inside the band with symptoms, the next step is a different test, methylmalonic acid. Inside the band without symptoms, there is no further test and the advice is to come back if symptoms appear. Inside the band with any of a short list of situations, including a condition that could deteriorate quickly, an irreversible cause such as autoimmune gastritis, certain surgery, or pregnancy and breastfeeding, NICE says consider replacement without waiting. Above the band, investigate something else, and if symptoms persist after 3 to 6 months consider repeating.

NICE is unusually candid about the limits of the whole exercise. There is no gold standard clinical or biochemical test for vitamin B12 deficiency, and no single widely adopted diagnostic algorithm. Four markers can be used, total B12, active B12, methylmalonic acid and homocysteine, and their accuracy varies between different groups of people. NICE also notes that of the 4 main manufacturers of total B12 tests, 3 have thresholds close to the ones it recommends and 1 does not.

That last sentence is the important one, and NICE builds an escape hatch around it

The recommendation carrying the table does not say to use 180 everywhere. It says to use the table, or, where there is substantial local variation in validated thresholds, to use the ones set by the lab doing the testing. That is not a footnote. It is an admission that the number is not transferable, and the UK data bear it out.

When UK labs taking part in the national quality scheme for these tests were surveyed in November 2024, the thresholds they were actually using to call a total B12 deficient ran from 65 to 239 nanograms per litre. That is a 3.7-fold range in what a UK lab calls deficient. Only 16.6% of them used NICE's 180. One of the major platforms was found to be incompatible with 180 altogether, with its own manufacturer recommending 145.

Reference intervals derived head-to-head on UK populations show the same thing from the other direction. On 3 platforms measured against each other, the ranges came out as roughly 190 to 678, 181 to 562, and 110 to 583 nanograms per litre. The lowest of those lower limits sits about 40% below the other 2. Applying one common cut-off across all 3 would label sufficient people on one platform as deficient.

There is an international standard, and it has not delivered

A World Health Organization international standard for serum B12 has existed since around 2007. Its own documentation says it will be re-evaluated when a reference measurement procedure has been established, which is a polite way of saying its assigned value is a consensus figure and not a traceable one. There is no reference measurement procedure for serum total B12 recognised by the international body that lists them. Manufacturers frequently calibrate against their own internal materials instead. When 5 platforms were compared against each other in 2025, one manufacturer's assays read consistently high and another's consistently low, and the agreement between them was too wide to draw a conclusion from. The finding was that the recommended thresholds should not be assumed to apply to all assays, and that this persists despite the international standard.

There is a small paradox in this, and it defeats the obvious workaround. You cannot infer a platform's bias from the reference range it prints. One manufacturer has the higher printed upper limits, and it is a different manufacturer's assays that read slightly high on quality assessment.

The interference that makes a normal result mean nothing, in the people who most need it to mean something

Routine total B12 tests work by competition, and the binding protein they use to do it is intrinsic factor, the same stomach protein that carries B12 through the gut. In pernicious anaemia the immune system makes antibodies against intrinsic factor. Those antibodies can bind the intrinsic factor in the test reagent, and because the signal in a competitive test runs backwards, the result comes out falsely normal or falsely high in exactly the person who has the condition.

This is documented across several analyser platforms. In published cases people with megaloblastic bone marrows and very high antibody titres have had B12 results sitting comfortably in range or above it, and 1 had a bone marrow biopsy that turned out to be unnecessary. In another, 3 cycles of plasma exchange were given before the answer emerged.

The honest position needs 3 more facts, and they cut the other way. Manufacturers know about this and build a defence into the assay. One denatures the sample first to inactivate the blocking antibodies. Another states in its method sheet that the assay is designed to avoid this interference, and in the published cases where one platform failed, that one agreed with the reference method. And when the question was tested systematically, taking antibody-positive and antibody-negative serum pools across 5 automated platforms and stripping the immunoglobulins out chemically to see whether the answer changed, no interference was found on any of them. One of the authors of that negative study had co-written an earlier case report, which is a scientist revising his own position downward.

So we could not find a reliable measurement of how often this happens. A figure of over 10% of pernicious anaemia cases circulates and we could not trace it to anything checkable. A figure of 50% also circulates and is a misreading of a statement about how often serum B12 is wrong for any reason. The defensible statement is that the interference is real, mechanistically understood, documented repeatedly in case reports, apparently a failure of the assay's antibody defence in people with unusually high or unusual antibodies, and of unknown frequency. NG239 does not address it, which is a genuine gap in the guideline.

The practical response published in the UK literature is the one worth carrying. Where there is strong clinical suspicion of deficiency and a normal B12, a functional test, homocysteine or methylmalonic acid, is what settles it. Neither is a test we offer, and neither is something to buy on a hunch.

Biotin. Total B12 is one of the markers biotin affects, and on this test it reads falsely high. Biotin, at the top of this guide, has the detail. One number belongs here. The manufacturer tested this assay to 50 nanograms per millilitre of biotin, and published peak serum concentrations after an ordinary 5 to 10 milligram hair, skin and nails capsule straddle that figure, with one source putting a 10 milligram dose at 55 to 140 nanograms per millilitre. So the tested tolerance may not cover what these products put in your blood, and the 8-hour wait the manufacturer specifies is not a precaution, it is the defence. In fairness to the assay, when that manufacturer's menu was characterised for biotin susceptibility, B12 was not among the most sensitive tests, with troponin, TSH and the thyroid antibodies far more affected. At 500 nanograms per millilitre and above, every assay tested showed significant interference.

How much does a repeat have to move

The top of this guide covers the reference change value. For total B12 it is about 20%. So 180 to 210 is noise and 180 to 260 is real.

The more interesting figure is the index of individuality, which compares how much your own B12 wanders with how much B12 varies between people. For total B12 it comes out at about 0.3, meaning your own range is roughly a third as wide as the population's. The consequence is concrete. Someone whose own settled level is 600 can fall to 300, which is a halving, and still sit inside the population reference range printed on a report. Note where 300 actually lands on the guideline, which is inside NICE's indeterminate band and not below it. A B12 in range is a weaker reassurance than it looks, and your own previous result is a far better comparator than the range printed beside it.

Metformin, and a regulator that changed its mind about how common this is

Metformin lowers B12. Measured in a placebo-controlled trial running over 4 years, in people with type 2 diabetes who were already on insulin, it fell by about 19%, and the risk of a B12 below 150 picomoles per litre rose by about 7 percentage points over that period.

In June 2022 the MHRA classified reduced B12 on metformin as a common side effect, meaning it may affect up to 1 in 10 people, and issued advice to test where deficiency is suspected and to consider periodic monitoring in people with risk factors. The risk factors it names are higher dose, longer duration, a starting B12 in the low-normal range, gut conditions affecting absorption, a vegetarian or vegan diet, an inherited predisposition, and taking a proton pump inhibitor or colchicine alongside. NG239 lists metformin as a risk factor. The monitoring advice is the regulator's, not NICE's.

Proton pump inhibitors and H2 blockers have a smaller and less well characterised effect in the same direction. Two or more years of a proton pump inhibitor is associated with a raised chance of a coded B12 deficiency, with a dose-response, though the study measuring that counted diagnoses instead of measuring the shift in the number, so it partly measures testing behaviour.

Pregnancy takes about half of it away, and nothing has happened

Total B12 falls markedly across pregnancy, by something like half. The reason is specific and it matters. Following healthy pregnant women through, B12 absorption does not change, the active fraction bound to the delivering carrier does not change, and the fall is entirely in the haptocorrin-bound pool, because the body makes less haptocorrin. Methylmalonic acid and homocysteine do rise.

So a pregnant woman with perfectly adequate B12 can read inside or below NICE's indeterminate band on a total B12. This is why NG239 specifies active B12 as the initial test in pregnancy, and why pregnancy is on its list of reasons to treat an indeterminate result instead of watching it.

The other things that move the number without moving your status

The combined pill. NICE puts this in writing, that the combined oral contraceptive can lower total B12 without causing a deficiency, while adding that a low result in someone taking it may still mean deficiency. The measured gap in young women is substantial, with users around 172 and non-users around 318 picomoles per litre. The reason to take NICE's framing seriously is that the functional markers do not follow the number down.

Anything you have already taken. NICE says to ask what over-the-counter B12 anyone is using, tablets, injections or patches, and to interpret with caution, because a supplement can raise the number without fully treating the deficiency. For injections the timescale is long. The UK product information for hydroxocobalamin states that 30% of a 1 milligram dose is retained, a range believed sufficient for body requirements for 2 to 10 months, which is why maintenance is every 2 to 3 months. NICE's instruction is blunt, which is not to repeat the diagnostic test at all in anyone having intramuscular replacement. There is one further trap. A B12 injection in the previous 1 to 2 weeks can also cause a false positive on the intrinsic factor antibody test, so recent treatment corrupts both tests, in opposite directions.

Nitrous oxide. Neither this test nor active B12 is suitable for anyone using nitrous oxide recreationally, and the reason is set out in the next entry.

Ethnicity. NG239 carries a recommendation that people of Black ethnicity may have a higher reference range for serum B12 than people of White or Asian ethnicity. The largest UK dataset on the question bears that out, with a higher interval in Black patients on the same analyser and no difference at all between Asian and White patients, so a single interval covers those 2. The published difference runs in one direction and is not a general rule that every group needs its own range.

Diet is the risk, and ethnicity is a poor stand-in for it. Vegetarianism among South Asian people in the UK is almost entirely an Indian pattern. In the same large UK cohort used for the vitamin D figures above, 27% of Indian participants were vegetarian against under 1% of Pakistani and Bangladeshi participants. Writing that South Asian people are often vegetarian is wrong for 2 of the 3 largest groups, and a B12 result is better read against what a person eats than against where their family came from.

A high total B12 in someone not supplementing

It is a result to notice and not to ignore. It can accompany liver conditions and some blood disorders, because the carrier proteins that hold B12 are released by those tissues. There is also a separate cause that is purely a measurement problem, where immune complexes form between immunoglobulins and the B12 binding protein and inflate the result. That one has been described as capable of masking a deficiency, and in the largest series looking for exactly that combination it was not found once.

The association people usually mean when they raise a high B12 is with cancer. It is real and it is strong. In a Danish registry study of people who had been referred for a B12 measurement, with anyone on B12 treatment excluded, a result between 601 and 800 picomoles per litre carried about 3 times the expected cancer incidence in the following year, and above 800 about 6 times, concentrated in blood cancers and in smoking and alcohol-related cancers. Work done elsewhere found that a raised B12 confirmed on a second measurement carried a real excess of solid cancers, and one that didn't persist had no such association.

No guideline we could find recommends acting on it. NG239 contains no recommendation about a high B12. The figures above come from registry populations, where people get tested because something prompted it, which inflates the association. The only step the primary work itself suggests is a repeat measurement to see whether it persists. That is not a test we would encourage anyone to buy for that reason, and it is not something this page can interpret for you.

Two standard warnings, and they do not hold up equally

Haemolysis is routinely listed as invalidating a B12 result, and the 2 method sheets are a long way apart on it. The most widely used active B12 assay states no interference up to a free haemoglobin of 1g/dL, which is obviously red cell haemolysis. The most widely used total B12 assay is far tighter, claiming no interference only up to 40 milligrams per decilitre and falsely low results above that, a limit its manufacturer tightened in 2024. We were not able to identify a study showing clinically significant haemolysis bias in a modern automated B12 assay.

Light protection is the other one. B12 is a photolabile molecule in a bottle, and the instruction to wrap the tube in foil circulates widely. We could not find a light protection instruction in the total B12 method sheet or the active B12 method sheets, and the total B12 sheet claims stability for 56 days frozen, which is hard to reconcile with meaningful light damage under ordinary handling. Light is good news for a posted sample but haemolysis is the one to keep an eye on, though the markers on this panel that genuinely cannot survive the post are homocysteine and potassium, not these.

Units

UK labs report B12 in either nanograms per litre or picomoles per litre depending on the analyser. Multiply nanograms per litre by 0.738 to get picomoles per litre, and picomoles per litre by 1.355 to go back. NICE's own table uses 0.738 and is internally consistent with it. Nanograms per litre and picograms per millilitre are the same number with 2 different labels.

One conversion error is worth naming because it is everywhere. The old conventional cut-off was 200 nanograms per litre. Quoted as "200 picomoles per litre" it becomes 271 nanograms per litre, which is well inside NICE's indeterminate band and not at the bottom of it. That single slip moves the line by 35%. NICE's 133 picomoles per litre and the historic 148 picomoles per litre are 180 and 200 nanograms per litre respectively, and they are not the same rule.

There is a pedantic point underneath the conversion that takes 1 sentence. The factor assumes all the B12 in your blood is cyanocobalamin, and it is mostly methylcobalamin and adenosylcobalamin, which have different molecular weights. Using 0.738 is correct practice, because that is how the assays are calibrated, but a picomole figure is best read as a cyanocobalamin equivalent.

Active Vitamin B12

Holotranscobalamin

The fraction of B12 bound to transcobalamin, the carrier that actually delivers it into cells. This is the portion available for use, and published figures put it at about 20% to 25% of the total.

It typically falls before total B12 does, so it detects developing deficiency earlier. It is the better test where symptoms suggest deficiency but total B12 has come back normal.

How much better is a fair question and the answer is modest. Its own proponents describe the improvement over total B12 as marginal to moderate. NICE treats total B12 and active B12 as equally acceptable starting points for most people and specifies active B12 only in pregnancy. The comparisons that put it ahead measure both against methylmalonic acid, which is itself imperfect as a reference. So the case for it is mechanistic and it is sound, but claims of clear superiority are not supported.

Where the numbers come from

NICE guideline NG239 gives the same kind of indeterminate band as it does for total B12. Below 25 picomoles per litre is confirmed deficiency, 25 to 70 is where the test has not settled the question, and above 70 deficiency is unlikely. Where symptoms are present and the result is indeterminate, methylmalonic acid is the next step.

Those 2 numbers belong to a particular assay, and this matters more here than it does for total B12

NICE's 25 and 70 come out of practice built on one manufacturer's assay. Published reference intervals for the active B12 assays in use run roughly 19 to 134, 21 to 123, 29 to 169 and 37 to 188 picomoles per litre. The last of those is a widely used automated platform, and when it was evaluated head-to-head against the assay NICE's figures derive from, it read about 9 picomoles per litre higher overall and up to 23 higher around the decision point where it actually matters. The people who measured that concluded the current cut-off values are not suitable for it and that it needs clinical validation before being used as a first-line screening test.

Put plainly, someone at 30 picomoles per litre on one platform, which NICE's table calls indeterminate, could be at 10 to 20 on another, which NICE's table calls confirmed deficiency. Which analyser produced the number is not a detail.

The reason this is worse for active B12 than for total B12 is slightly perverse. Active B12 assays are actually better harmonised on paper, because they share a common primary calibrator held in Dundee. Sharing a calibrator has not made the results interchangeable.

It is a noisier repeat measurement than total B12, not a steadier one

This cuts against how active B12 is usually sold, including by us, so it is worth stating clearly. Measured in healthy people sampled weekly, the within-person variation for active B12 is roughly double that of total B12, and the reference change value comes out at about 37% against about 20%. Two active B12 results have to differ by well over a third before the difference means anything.

Its advantage is biological relevance, not reproducibility. It is measuring the right fraction. It is not measuring it more precisely.

What else moves it

It is not supplement-proof, and if anything the opposite. Because it is the fraction newly absorbed B12 binds to first, it responds faster to a recent dose in both directions than total B12 does. NICE's caution about interpreting results in people already taking an over-the-counter preparation is written to cover total or active B12, not just total.

Reduced kidney function raises it, and how far down the range that matters is genuinely unsettled. In dialysis and end-stage kidney disease active B12 runs markedly high alongside markedly raised methylmalonic acid, and one group that measured it in renal patients concluded outright that active B12 cannot be used as a marker of B12 status in renal dysfunction. Against that, the largest study of older men with mildly to moderately reduced kidney function found no meaningful effect. Both may be right about their own populations. What is not in dispute is that methylmalonic acid also becomes unreliable as kidney function falls, which matters because it is the test NICE sends you to next.

Inherited variation in the carrier itself shifts the number a little. The version of this claim in circulation names the wrong variant. The one with real evidence behind it, tested in over 2,000 people, moves active B12 by around 14 picomoles per litre between the 2 homozygous forms while leaving total B12, methylmalonic acid and homocysteine untouched. That is a good illustration of the general problem with population ranges. It moves where you sit in the range without moving anything about your B12 status.

Biotin. Active B12 is one of the markers biotin affects, and on this test it reads falsely low, because it is built the opposite way round from the total B12 test. Biotin, at the top of this guide, has the detail. The tested tolerance for this one is 40 nanograms per millilitre, which is lower than total B12's 50, and published peaks after an ordinary hair, skin and nails capsule straddle both.

The intrinsic factor antibody problem should not apply here

The previous entry describes how antibodies against intrinsic factor can make a total B12 falsely normal in pernicious anaemia. Active B12 assays do not use intrinsic factor anywhere in the reagent chain. They use an antibody against the delivering carrier to capture it and a second antibody to detect it, so the mechanism that corrupts a total B12 has nothing to attach to.

That follows from the assay design and it is stated in review articles. We could not find a published interference study demonstrating it directly, and none of the active B12 method sheets we read makes a claim about intrinsic factor antibodies either way. Active B12 is also not immune in principle to other antibody interferences. So the honest version is that this particular problem should not reach active B12, for a good structural reason, and that nobody has published the experiment.

And one situation where neither B12 test is the right one

Where vitamin B12 deficiency is suspected because of recreational nitrous oxide use, NICE says the initial test should be plasma homocysteine or serum methylmalonic acid rather than either B12 measurement. NICE's stated reason is that nitrous oxide inactivates the B12 molecule, so total or active concentrations may appear normal even when there is a deficiency.

The biochemistry is slightly more specific than the shorthand, and the specifics are the reason active B12 does not help. Nitrous oxide attacks B12 while it is sitting in the active site of methionine synthase, one of the 2 enzymes B12 serves. The reaction generates a reactive oxidant that damages the enzyme, and the oxidised B12 falls off. So the enzyme is destroyed and the B12 circulating in the blood, which is what any B12 test measures, is untouched in quantity. Measuring the delivering fraction instead of the total does not fix it, because the problem is not which carrier the B12 is riding on.

The order in which things then move matters. Homocysteine rises first and rises most, because methionine synthase is the enzyme that clears it and it is the enzyme being destroyed. Serum B12 and active B12 fall afterwards. A B12 result can be normal at presentation, which is the whole reason NICE sends you elsewhere, though in published case series it is more often low or low-normal by the time people present, and methylmalonic acid is usually raised alongside it.

Neither of the 2 tests NICE asks for is a test we offer, and homocysteine cannot be offered by post at all, because it has to be separated from the cells within about an hour and kept cold before it is. So anyone in this situation cannot be assessed by anything we sell, and it is better to say so than to sell you the wrong one.

Folate

Vitamin B9, serum folate

Vitamin B9, essential for making DNA and for red blood cell production. Deficiency produces a similar picture in the blood count to B12 deficiency, large red cells and anaemia, which is why the two are usually measured together. Why they produce the same picture, and why folate can correct that picture while a B12 problem carries on underneath it, is in Nutrients, in the Body Guide.

In pregnancy folate deficiency raises the risk of neural tube defects, which is why supplementation is advised before conception.

Serum folate reflects recent intake and not long-term stores, so it can rise after a single folate-rich meal. That is the shortest memory of anything on this panel and it shapes everything below.

Something is about to change about this marker, and part of it has already happened

From 13 December 2026, fortifying non-wholemeal wheat flour with folic acid becomes a legal requirement across the UK. The level set is 250 micrograms per 100 grams of flour, and it is what has to be added rather than what survives processing. The 2019 impact assessment modelled a 25% loss between fortification and consumption when it estimated the resulting rise in intake.

The part that matters for anyone testing now is that some flour changed before the law did. Nothing prevents a miller fortifying early, and Food Standards Scotland said in October 2025 that fortification of some products was expected to start in autumn 2025. We have no basis for saying what proportion of UK flour is fortified today, and that is rather the point. A result taken now is being taken during a gradual and unevenly distributed change, which is harder to read than a clean switch would be, because you cannot tell from a number whether the bread that person eats was fortified.

What is and is not covered is narrower than people assume. It applies to non-wholemeal flour milled from common wheat, Triticum aestivum, milled in or imported into the UK. Wholemeal is exempt, and so is flour for communion wafers, matzos, gluten and starch. A small mill is exempt if its site has a maximum annual capacity of 500 tonnes and it produced under 500 tonnes in each of the last 3 calendar years, which is a 2-part test and not simply a size. Gluten-free and non-wheat flours fall outside it, as do other wheat species including spelt and durum, because the duty names only common wheat. And imported products made of multiple ingredients are not covered, which means imported bread, biscuits and pasta are outside it.

Two details are worth having because the policy did not follow the advice. The Scientific Advisory Committee on Nutrition recommended 300 micrograms per 100 grams, a figure it first gave in 2006 and restated in its 2017 update, not the 250 adopted. And it said mandatory fortification should only be introduced alongside restrictions on voluntary fortification of other foods, to keep intakes below the 1 milligram a day guidance level. No such restriction forms part of the 2026 regulations, so fortified cereals, spreads and supplements continue unrestricted on top of the new flour. The government's own expectation is about a 20% reduction in neural tube defects, and there is a published argument that the level chosen is far too low to do much more than that.

What that does to reading a folate result

Three things, and all of them are about comparison and not about health. A serum folate of 12 in 2024 and a serum folate of 12 in 2028 are not the same finding, so any comparison of a result across the transition is comparing incommensurable numbers. A low result becomes rarer and less specific, because in a fortified population it points to absorption, alcohol, drugs or a sampling problem far more often than to diet. And the high end becomes common and stays meaningless, because there is no established upper reference limit for serum folate with a clinical meaning behind it.

For scale, the national diet and nutrition survey found 52% of UK adults aged 19 to 64 below 13 nanomoles per litre, and 11% below 7, in its most recent combined years, with geometric mean serum folate falling significantly across most age and sex groups over the preceding decade. Countries with mandatory fortification sit far higher. The UK programme is narrower in reach than the American one, which covers a wider range of grain products.

Where the numbers come from, and the honest answer is nowhere

There is no single agreed threshold, and the position is worse than that sounds. The UK haematology society published one in 2014, putting deficiency below 7 nanomoles per litre with an explicit indeterminate zone of about 7 to 10. It has since marked that guideline as superseded by NICE NG239. NG239 sets no folate threshold and gives no guidance on diagnosing or managing folate deficiency, its scope being B12. Folate appears in it only as something that can raise homocysteine and therefore confuse a B12 result. So the only UK folate threshold that ever carried a formal grade of recommendation now sits inside a superseded document, and nothing graded has replaced it.

The World Health Organization publishes 2 different figures, not 1, and which you get depends on which harm you anchor to. Anchored to macrocytic anaemia it is below 6.8 nanomoles per litre. Anchored to a raised homocysteine it is below 10. Those are 47% apart for the same word.

The College of American Pathologists takes a different approach entirely. Rather than publishing a number, its test-ordering guidance restricts when folate should be measured at all, saying it should be measured only where the history supports a deficient diet or a condition affecting absorption, and that it is not necessary to measure it in people with adequate diets and no evidence of malabsorption. That is a notable position for us to be quoting, and it is the honest one to quote.

Reference ranges for serum folate differ between labs more than for most markers, so the range printed on your own report is the one that applies. How much the line matters was shown neatly across 2 large American hospital populations. In the same people, 0.6% fell below 3 nanograms per millilitre, 1.6% below 4, 4.9% below 5.5, and 43.2% below 13. Nothing about anyone changed. Only the line did.

The assay moved under everyone's feet, twice, and it is still moving

This is the part of the folate story that a guide is obliged to tell, because it explains results that otherwise look like a change in the person.

When the most widely used manufacturer recalibrated its folate assay to the international standard, UK results fell, with a median negative shift of about 21% and much larger shifts at low concentrations, which is exactly where the decisions get made. On the same population, the proportion classified deficient went from 3% to 18%.

A new version of that manufacturer's third-generation folate assay was introduced in October 2023 and adopted through 2024. The UK quality assessment scheme has documented a lot-dependent negative bias in it from September 2024 onwards, tracked by reagent batch, with the method means for that manufacturer sitting below the all-lab mean for every specimen. In 2 Swedish regions running the same assay, the median patient folate fell by 19% between early 2023 and late 2024, the proportion of results below 7 rose from about 4% to about 15%, and the proportion below the manufacturer's own lower limit of 8.83 rose from 11% to 25%. About 40% of the UK labs taking part in that quality scheme run that platform.

Set that beside the fortification change and the difficulty becomes clear. Fortification is pushing folate results up and an assay drift is pushing them down, over the same 3 years. Anyone drawing a conclusion from a trend in UK folate results between 2024 and 2027 without accounting for both is going to get it wrong, and a Northern Ireland hospital biochemistry department published a letter in July 2026 asking precisely that question, whether an apparent improvement in folate status there was fortification or the assay.

Between methods, the disagreement is large in absolute terms. Measuring the same international standard, 4 platforms recovered between 91.7% and 109.0% of it, a spread of about 19 percentage points. Measured against the mass-spectrometry-certified reference material, the biases ran from about 9% low to about 26% high, a spread of 34 points. The desirable analytical goal is 3%.

One thing we are not going to claim, because we went looking for it and the evidence does not support it. Folate is often described as the worst-standardised of the common blood tests. When 5 common blood tests were compared across 38 labs on fresh samples in a Norwegian quality scheme, folate was the only one of the 5 that met its bias goal. Both things are true, and the reconciliation is that folate's allowable bias is generously wide because its natural variation is wide. The accurate statement is that folate has large between-method disagreement and incomplete standardisation, not that it is the worst.

Haemolysis, and this is the one that matters most for a posted sample

Red cells contain far more folate than serum does. So when red cells break, folate leaks out and the serum result goes up. The manufacturer of the most widely used assay does not describe this as a limitation. Its own wording is that haemolysis may significantly increase folate values and that haemolysed serum samples are therefore not suitable for use in the assay. Other reference labs reject haemolysed samples outright.

The direction is what makes this serious. Haemolysis pushes folate up, so the failure mode is a falsely normal or falsely high result in someone who is genuinely low. It is a false reassurance and not a false alarm. And unlike potassium, where a haemolysed sample produces an obviously impossible number that gets caught, an inflated folate lands inside the reference interval and looks entirely unremarkable. Folate is one of a small group of tests, with LDH, AST, direct bilirubin and potassium, showing the highest degree of haemolysis interference of the 42 that group tested.

Pulling in the other direction, folate degrades. It falls by 10% to 20% after more than 8 hours of light exposure, and it is limited to about 8 hours at room temperature before degradation starts to matter. A postal sample sitting warm and lit is losing folate while any haemolysis in the same tube is adding it, and the 2 errors do not cancel in any predictable way.

Venous samples only

Folate is accredited for venous blood and is not reported on a finger-prick sample.

Fasting, where the labs genuinely disagree

There is no national or professional guidance on fasting for folate, and practice splits. One major reference lab requires an 8-hour fast and states plainly that non-fasting specimens yield falsely elevated results. UK hospital handbooks frequently specify no fasting requirement at all. Nobody adjudicates.

The underlying biology is not in dispute. At steady state, serum folate rises by about 11.6% for every extra 100 micrograms a day of folic acid, and takes about 13 weeks to get there. In the other direction, a single dose above roughly 200 to 250 micrograms puts unmetabolised folic acid into the blood within one meal. So a supplement or a fortified breakfast measurably changes circulating folate the same day. What we could not find anywhere is a clean figure for how big that single-dose bump is and how long it lasts, and we are not going to invent one.

How much does a repeat have to move

About a third. Published biological variation figures put the reference change value for serum folate at about 32%, so 12 to 15 is not a change. That figure comes from one assay in one lab, and the real-world critical difference across a platform change or a reagent lot change is larger, for the reasons above.

The index of individuality for folate works out at about 0.55 from the same figures. That is below the conventional line, so a person's own previous result is the better comparator, but folate came out as much the less individual of the 2 markers. Population ranges fit folate better than they fit B12, and the individuality argument should not be overstated for this one.

Red cell folate

Red cell folate reflects longer-term stores instead of recent intake, and it is not one we offer. That is the usual reason given for wanting it, so it is fair to add that the professional bodies advise against it. The UK haematology guideline carried its strongest grade of recommendation against routine red cell folate testing, reserving it for strong clinical suspicion despite a normal serum folate and having already ruled out a B12 problem, which it estimated covers about 5% of people with large red cells and a normal serum folate. The American pathology body lists it among tests that are rarely indicated and states that it generally adds no diagnostic value over a serum folate. It is also analytically worse, with more variation between methods than serum folate, because the red cells have to be broken open first. It takes about 36 weeks to reach a steady state after a change in intake, against 13 for serum.

Drugs, alcohol, and the results that should not have been reported

Methotrexate and folinic acid are folate analogues, and both bind the folate binding protein the test is built on, so they can cross-react with it. Reference labs list samples from patients taking them as contraindicated for this assay.

Alcohol depresses serum folate quickly and it recovers quickly. This was worked out carefully a long time ago. Oral alcohol on a low-folate diet dropped serum folate further than the diet alone, largely within the first day. An alcohol infusion produced a striking fall between the 8th and 10th hour that returned rapidly to normal once it stopped. In 2 people followed over weeks, serum folate fell with alcohol, rose when it was stopped and fell again when it was resumed. A heavy session shortly before the sample is a real reason for a borderline low result.

The UK haematology guideline's own list of things that produce a falsely low serum folate is short and useful. Anorexia, acute alcohol consumption, normal pregnancy, and anticonvulsant therapy. Note its framing. The number really is low. The conclusion "folate deficient" is what is wrong.

Biotin. Folate is one of the markers biotin affects, and Biotin at the top of this guide has the detail. Two things are specific to this one. Its tested tolerance is 21 nanograms per millilitre, which one UK hospital's summary of biotin interference puts below every other assay on that analyser's menu, and well under what an ordinary hair, skin and nails capsule produces. And the direction is less simple than the general rule. At high biotin concentrations it reads high as expected, and worst when the folate itself is low, but just above the tested limit a reproducible bias of about 10% in the other direction has been measured. There is also a disagreement about how long to leave it. The manufacturer says at least 8 hours after doses above 5 milligrams a day. At least one major lab running the same assay tells people to stop biotin 72 hours beforehand. If you take it, the conservative instruction is the defensible one.

The folate and B12 question, stated as it actually stands

That folate can correct the blood picture of a B12 deficiency, removing the clue while nerve damage carries on, is the reason the 2 are read together. Nutrients, in the Body Guide sets out the mechanism and the history, including that the stronger claim, that folate actively accelerates the nerve damage, rests on uncontrolled case series from the 1940s and 1950s at doses of 5 to 20 milligrams a day, that 2 separate reviews have concluded those cases do not establish causation, and that a respected neurologist maintains they do.

Two additions belong here and not there, because they are about the numbers. First, the current UK product information for folic acid 5 milligram tablets does state it, in terms, saying folic acid alone will not prevent and may precipitate subacute combined degeneration of the cord in pernicious anaemia or undiagnosed megaloblastic anaemia. That is a live regulatory position at that dose. Second, the separate and much-repeated claim that a high folate makes things worse in someone with a low B12 is disputed and not established. The supporting work defines high folate as above 59 nanomoles per litre, which is several times the typical UK adult level, its own authors concede the effect was absent before American fortification, and a UK study of 2,403 people looking for the same interaction at the same cut-off found nothing.

The one conclusion nobody disputes is the useful one. Do not use the blood count as the trigger for suspecting a B12 problem. Measure B12.

Pregnancy, where the test has no role

UK guidance is 400 micrograms of folic acid daily from before conception to the end of the twelfth week. A higher dose of 5 milligrams daily applies to a list of higher-risk groups including a previous pregnancy affected by a neural tube defect, diabetes, sickle cell anaemia, thalassaemia, epilepsy treated with anti-epileptic medication, and medicines affecting folate absorption. The duration differs by group, and in some, sickle cell anaemia and thalassaemia among them, the higher dose is usually continued through the pregnancy rather than stopping at 12 weeks, so this is a conversation to have with whoever is looking after the pregnancy.

UK guidance on folic acid in pregnancy is built around taking a fixed dose and not around measuring anything, and we could not find a UK guideline that recommends testing folate before or during pregnancy. The reasoning holds up. The dose is fixed regardless of the result, so a measurement cannot change the advice. Serum folate takes about 13 weeks to reach a steady state and red cell folate about 36, so neither can confirm adequacy on the timescale of a decision about conceiving. And normal pregnancy is on the list of causes of a falsely low serum folate, so a flagged result in pregnancy is manufacturing anxiety. Take the dose. Do not test to decide.

The evidence behind the 5 milligram figure is thinner than its confident use suggests. When NICE looked for trials of high-dose folic acid in pregnancy it found none at the recommended dose, had to relax its criteria to admit 4 milligram studies, and rated the evidence it did find as low to very low. The 5 milligram recommendation is expert consensus and not trial evidence.

MTHFR

There is a large amount of consumer content built on MTHFR genotyping and folate. Two professional bodies recommend against the test, and one of them is British. The UK haematology society's thrombophilia guideline states that genetic testing for variants without a clinically significant link to thrombosis, naming MTHFR, is not recommended, and that MTHFR testing should not be in thrombophilia panels unless features of homocystinuria are present. The American medical genetics body reached the same conclusion, finding that meta-analyses had disproven the association with venous thromboembolism and that in people homozygous for the common variant with a normal homocysteine there is no evidence of increased risk. Its document was later reclassified from a practice guideline to a clinical practice resource, on the stated grounds that it no longer met the college's criteria for an evidence-based practice guideline. That is a downgrade of the document, not a reversal of what it says, and it is sometimes cited as though it were the latter.

The common variant is common, and the homozygous form modestly raises homocysteine. That is where the established part stops.

Units

UK labs report serum folate in nanomoles per litre or in micrograms per litre, which is the same as nanograms per millilitre. Multiply micrograms per litre by 2.266 to get nanomoles per litre. So 3 micrograms per litre is 6.8 nanomoles per litre and 7 nanomoles per litre is about 3.1 micrograms per litre.

Magnesium

A mineral involved in several hundred enzyme reactions, including energy production, muscle and nerve function, and blood pressure regulation. Why the list is that long, and what magnesium has to do with parathyroid hormone and calcium, is in Minerals and bone, in the Body Guide.

Serum magnesium has a real limitation. Less than 1% of the body's magnesium is outside cells at all, and less than 0.3% is in serum. The rest is in bone and inside cells. The body defends the blood level tightly, so serum magnesium can stay normal while tissue stores are depleted. A low result is meaningful. A normal one does not rule depletion out.

Genuinely high serum magnesium is uncommon and almost always involves either reduced kidney function or large doses of magnesium-containing supplements or laxatives. Of people found with a raised magnesium, over 70% had a reduced filtration rate and 85% had chronic kidney disease. Magnesium oxide, a common laxative, and Epsom salts are the usual medicinal routes.

How weak is the test, in numbers

The compartment argument above is unanswerable in principle. If 99% of your magnesium is somewhere a blood test cannot see, a blood test can be normal while that 99% is not. The direct evidence is scarcer than the confidence of the claim suggests, and the best of it is worth reporting in full because it cuts both ways.

The nearest thing the field has to a reference method is a magnesium loading test, where magnesium is given and the amount retained rather than passed out is measured, on the reasoning that a depleted body holds onto it. Measured against that, in over 100 intensive care patients and 41 healthy people, serum magnesium explained only about a tenth of the variation in how much was retained. That is damning. What is usually left out is the rest of the same finding, which is that serum magnesium was the only measurement that predicted retention at all. Red cell magnesium and urinary magnesium predicted nothing.

So the popular conclusion, that red cell magnesium is the test serum magnesium failed to be, is not supported and is contradicted by the one direct head-to-head. Minerals and bone, in the Body Guide covers why it is sold anyway. There is no validated test of individual magnesium status. Serum magnesium is the only practical option, it is poor, and every proposed alternative is either unvalidated, unavailable, or has been shown to perform no better.

Where the numbers come from

We could not find a professional body publishing a threshold for a healthy person. UK NHS documents in current use put hypomagnesaemia below 0.7 millimoles per litre and give normal ranges of 0.7 to 1.0 or 0.7 to 1.1. The conventional 0.75 to 0.95 interval that appears in the literature comes from an American population survey carried out between 1971 and 1974, and it is a description of where people sat and not a line drawn from outcomes. The population reference range is assay-dependent, and between UK sources the upper limit alone varies by about 15%, so a result of 1.05 is normal at one lab and high at another.

A figure of 0.85 circulates and it is not a deficiency threshold. It has been argued in the nutrition literature, by groups whose work is specifically on magnesium, that the lower limit should move to 0.85 millimoles per litre, on the reasoning that current intervals were built from populations quietly containing people who were depleted. Two things before anyone uses it. In a recent American survey about 68% of adults fell below 0.85, and a line that calls two-thirds of the population deficient is a risk gradient and not a diagnostic threshold. And it has not been adopted into UK practice, with every UK document above still using 0.7. The legitimate half of their argument does survive, which is that the conventional interval really is a 1970s population distribution and not an outcome-validated threshold, and that the variation between labs really is large.

The population range is the wrong comparator here, and this one is measurable

The index of individuality compares how much your own magnesium moves against how much magnesium varies between people. For magnesium it comes out at about 0.56, which is below the conventional line of 0.6. That means your own set point occupies a narrow slice of the population range, and you can move a long way from your own normal while sitting comfortably inside 0.7 to 1.0. You are being compared with the population when you should be compared with yourself. Note that this is a purely statistical argument and it reaches the same conclusion as the compartment physiology above by an entirely independent route.

The reference change value follows from the same figures and comes out at about 11%, which is roughly 0.09 to 0.10 millimoles per litre. A UK source arrived at the same place, stating that changes of the order of 0.1 must occur before genuine change can be confirmed. So a magnesium moving from 0.78 to 0.85 between 2 tests has not moved.

Everything that goes wrong with the sample pushes this number up

It is the opposite of what people expect.

Time to separation is the big one. Magnesium passes out of red cells into the serum around them, so what matters is how long the serum sits on the cells. Separated serum is stable for about a week refrigerated. Unseparated whole blood kept cool starts to drift after about 4 days. Unseparated whole blood at room temperature is out of specification after a single day and keeps rising. Published advice is to centrifuge within an hour.

Posture and tourniquet both add to it. Going from sitting to standing raises magnesium measurably, along with albumin, calcium and most of the protein-bound analytes, because standing shifts plasma volume. A tourniquet left on for 30 seconds changes nothing. At 60 seconds magnesium rises significantly, and at 2 to 3 minutes the rise is clinically significant.

Haemolysis raises it, and by much less than the warnings imply. The live version of this claim usually invokes the same mechanism as AST in the liver group. That comparison overstates it. Red cells hold only about 2 to 3 times as much magnesium as serum, against roughly 25 times for potassium and 10 times for zinc, so the gradient driving the artefact is small. Measured across bands of free haemoglobin, one study found magnesium stayed within acceptable limits at every level up to 4.5 grams per litre, another found an increase only from 5 grams per litre, which is frank and obviously red haemolysis, and some investigators have found no interference at all. Magnesium behaves nothing like potassium here. A faintly pink sample is a reason to doubt the potassium, and potassium is not on our menu for exactly that reason.

Contamination from the wrong tube is the one exception, and it runs downward. If a little EDTA from a full blood count tube gets into the serum tube, it grabs the magnesium and the colorimetric test cannot see it. The size of that effect is method-dependent and smaller than folklore suggests, and on at least one modern platform a modest degree of contamination moved magnesium by less than the reference change value. The giveaway is not the magnesium. It is the pattern. A high potassium, a low calcium, a low magnesium and a low alkaline phosphatase appearing together on one sample are 1 artefact, not 4 findings.

So the practical conclusion is unusual and it is the useful one. Every handling problem a posted sample is prone to pushes magnesium up. A low magnesium on an imperfectly handled sample is therefore more credible than a normal one.

Fasting, and the small daily rhythm

Fasting is not a requirement for serum magnesium. No guidance requires it and no method sheet we read imposes one, which is a sharp contrast with zinc in the next entry, where separate thresholds exist for fasted and unfasted samples. A small circadian variation has been described, but it is not large enough to change how a routine sample is read. Posture, tourniquet time and time to separation all matter considerably more.

The methods do not agree with each other either

The most recent published UK breakdown by method is old, from the national quality scheme in the mid-1990s, and it found within-batch variation running from 4.6% to 7.4% depending on which chemistry the lab used, against a desirable figure of 1.6% or better derived from magnesium's own biological variation. The assessment at the time was that magnesium assays fall well short of desirable performance in general practice, and we could not find a current published UK distribution to say whether that has changed. Assay failure is not hypothetical either. A 2026 report describes a magnesium bias incident on one analyser where 63% of results exceeded the allowable error threshold before the cause, a contaminated probe, was found.

Drugs, and one regulator has been explicit

Proton pump inhibitors are the important one. The MHRA issued advice in 2014 to consider measuring magnesium before starting a proton pump inhibitor and periodically during prolonged treatment, particularly in anyone also taking digoxin or a drug that lowers magnesium such as a diuretic. Some cases appeared after 3 months and most after a year. The American regulator's review found that in about a quarter of the cases it looked at, magnesium supplementation alone did not correct the level and the drug had to be stopped.

The other drugs UK hospital guidance lists as lowering magnesium are loop and thiazide diuretics, aminoglycoside antibiotics, amphotericin, ciclosporin, tacrolimus, theophylline, salbutamol, and platinum-based and several other cytotoxic drugs. Heavy alcohol use lowers it by reducing how much the kidney reabsorbs, and low magnesium is common in people with alcohol dependence even before any sign of liver injury.

Units

UK labs report magnesium in millimoles per litre. American reporting uses milligrams per decilitre and older obstetric practice uses milliequivalents per litre. Multiply millimoles per litre by 2.43 for milligrams per decilitre and by 2 for milliequivalents. So 0.70 is 1.70 milligrams per decilitre, 0.85 is 2.07, and 1.00 is 2.43.

The small conversion factor is itself a trap. Because the 2 common units differ by only about 2.4 times, a mislabelled result is not obviously absurd the way a mislabelled glucose would be. An American-format result of 2.0 milligrams per decilitre read as 2.0 millimoles per litre looks like a markedly raised magnesium when it is in fact slightly low.

Zinc

A mineral required for immune function, wound healing, protein synthesis, taste and smell, and testosterone production.

Serum zinc has a daily rhythm and falls after eating, so a morning sample before food and before supplements is the most consistent. It also falls during inflammation and acute illness independently of how much zinc is in the body, so a low result during infection is difficult to interpret.

This marker sits outside the lab's UKAS ISO 15189 scope. Same lab, same analysers. What that means in practice is set out on our Turnaround Times page.

Where the numbers come from, and what the bodies that publish them actually say the test is for

We could not find a UK professional body publishing a threshold for zinc. Neither the Scientific Advisory Committee on Nutrition nor NICE appears to have a position on serum zinc testing or cut-offs, and we looked for both.

The figures in general use come from an international zinc nutrition group, and they are unusual in a way that matters. They are stratified by time of day and by whether the person had eaten.

For men aged 10 and over, a morning fasting sample below 11.3 micromoles per litre, a morning non-fasting sample below 10.7, or an afternoon sample below 9.3 counts as low. For women aged 10 and over the equivalent figures are 10.7, about 10.1, and 9.0. In pregnancy they are 8.6 in the first trimester and 7.6 in the second and third. As far as we can find, serum zinc is the only routine nutritional blood test with published cut-offs that depend on the clock. That stratification is itself the strongest available admission that the conditions of sampling dominate the number.

These are statistical limits and not clinical ones. They are the bottom 2.5% of a reference population, which is a different thing from a level at which something goes wrong. Overt clinical signs of zinc deficiency track a considerably lower figure, around 7.6 micromoles per litre. The reference population they came from was American and was surveyed between 1976 and 1980. When the same exercise was repeated 45 years later on over 12,000 rigorously screened healthy Indian children and adolescents, the cut-offs came out 10 to 18 micrograms per decilitre lower, and applying them cut the estimated prevalence of deficiency by between 2.7 and 5.5 times.

The bodies that publish these figures say, repeatedly, that they are for populations

This is the part that a company selling the test has a duty to print. The international zinc group frames serum zinc as the best available biomarker for assessing the risk of zinc deficiency in target population groups, the justification being the absence of anything better and not positive validation of this one. The international expert panel on nutrition biomarkers lists its limitations in its own summary, noting it responds less to zinc in food than to a supplement taken between meals, that there is considerable variation between individuals, and that it is influenced by recent meals, time of day, inflammation, and certain drugs and hormones. The standard nutritional assessment reference states plainly that for an individual, serum zinc is not a reliable indicator of zinc status. A national micronutrient survey manual is blunter still, saying serum zinc cannot be used in the diagnosis and treatment of individuals. And the UK clinical biochemistry body's own public information says that general screening for zinc deficiency is not recommended.

The finding underneath all of that is the one people find hardest to believe. Serum zinc responds to intake, and only weakly. Pooling the controlled trials and observational studies in adults, doubling zinc intake shifts serum or plasma zinc by about 6%. In a survey of over 4,000 people, habitual zinc from diet or from supplements showed no detectable effect on serum zinc at all. Those 2 findings are both real and they are not in conflict. Serum zinc moves measurably when someone is experimentally depleted or supplemented, and it does not track what people ordinarily eat or where their own body stands. That is the reason the literature on it looks contradictory, and it is the reason a serum zinc is a poor guide to your own intake.

The most recent systematic review of the field assessed 72 candidate biomarkers of zinc status. Only serum or plasma zinc and urinary zinc excretion showed a dose-response at all, and its conclusion was that no clearly superior alternative emerged. Red cell zinc is not the rescue either. In a 49-day depletion study it did not change despite documented impairment of taste and immune function, and supplementation at 50 milligrams a day for up to 3 months produced no response. The expert panel classified it as not useful.

The time of day matters several times more than the fasting does

Both are real and the usual emphasis is backwards. Measured across more than 4,000 people, the average serum zinc was 86.8 micrograms per decilitre on a fasting morning sample, 85.0 on a morning sample after food, 76.9 in the afternoon and 73.8 in the evening. So fasting accounts for about 2% and the time of day for 9% to 15%. One review puts the possible within-day swing in an individual as high as a third.

The meal effect itself has been measured properly. In a study feeding identical meals 6 hours apart and sampling 8 times in between, the postprandial pattern accounted for half of all the within-person variation in plasma zinc, with a small rise of 2% to 6% in that study's first hour and then a decline to a low point about 4 hours after the meal. The pattern disappeared entirely during a fasting control. So the advice to take the sample in the morning is the one that matters, and taking it before food is a refinement on top of it.

The analytical window is wider than the decision

The UK quality assessment scheme for trace elements judges a serum zinc result acceptable if it lands within about 15%, or 1.2 micromoles per litre, of the target, whichever is larger. Put in customer terms, a serum zinc of 10.0 is, by the scheme's own acceptability criterion, indistinguishable from anything between roughly 8.8 and 11.2. That window straddles the cut-off.

The reason is not the type of instrument. When 7 labs in 4 countries ran the same samples on 9 instruments of 3 different technologies, the technology was not a significant determinant of accuracy or precision in that study. Calibration between individual instruments differed by up to 18.9%, including between instruments of the same type. And serum and plasma had the worst precision of any material tested, worse than liver tissue and worse than the control materials.

Everything that goes wrong with the sample pushes this number up too

The tube matters, and the famous figures are 40 years old. Zinc leaches out of rubber stoppers and clot activators. The 2 studies establishing that, from 1979 and 1983, reported ordinary evacuated tubes raising serum zinc several-fold, roughly tripling it in one and by around 250% in the other. Those figures describe hardware that has since changed and quoting them undated is misleading. Modern paired comparisons of a gel serum tube against a purpose-made trace element tube give about 6.8% higher on the gel tube, which is real and above the allowable bias, and an order of magnitude smaller than the old numbers. One study drawing paired samples found no systematic difference at all, with the caveat that it applies to the particular batch of tubes used. What the evidence implicates is the stopper formulation and the clot activator. We could not find a primary study isolating the gel itself, so the widely repeated claim that the gel absorbs zinc is asserting a mechanism nobody has demonstrated.

Haemolysis raises it. Red cells hold roughly 10 times as much zinc as serum, so lysing well under 1% of them raises serum zinc by about 5%, which corresponds to about 1 gram per litre of free haemoglobin in the sample. Reference labs reject at about that figure and above.

Delay to separation raises it. Zinc comes out of platelets over the first couple of hours, then keeps climbing. Delays of 4 to 24 hours before separating the serum raise plasma zinc by 5% to 12% and serum zinc by 0.5% to 7%, and that happens regardless of the temperature the sample was kept at. Published protocols call for separation within 30 to 40 minutes, and one major reference lab rejects a sample not separated within 2 hours.

A tourniquet raises it, by the same fluid-shift mechanism as magnesium, which is why the protocols restrict it to about a minute.

A finger-prick sample would raise it, and that is a good reason not to sell zinc that way. Drawn from the same people in the same session, capillary plasma zinc ran 8% higher than venous. The classification consequence is the striking part. In the same group of people, 28% counted as zinc-deficient on the capillary samples against 53% on the venous ones. This does not arise for our zinc, which is a venous-only add-on, but it applies if you have a capillary zinc from anywhere else.

The one thing that pushes it down is the wrong tube in the other direction. EDTA carried over from a full blood count tube collapses a colorimetric zinc result, by around 70%, which makes zinc one of the most sensitive indicators of that contamination. Zinc measured by mass spectrometry is unaffected by it.

Same conclusion as magnesium, for the same reason. Every handling problem pushes the number up. A low zinc on an imperfectly handled sample is more credible than a normal one.

Inflammation, and the rule is operational and not a guess

Zinc moves out of the blood and into the liver during an acute phase response, driven by the same cytokine signalling that raises CRP. The published rule comes from a UK trace element reference lab that stratified over 2,000 samples by CRP band, and it is that a reliable interpretation of a plasma zinc can only be made when the CRP is below 20 milligrams per litre. That is a measurable condition and not a waiting period, which is why it is better advice than "wait until you feel better". hs-CRP is on our menu and is the marker to read alongside it.

Two findings about inflammation and zinc pull against each other, and it is worth saying why. The rule above is about reading one person's result, where an ongoing acute phase response makes a low zinc uninterpretable. The findings below are about what inflammation does to zinc across whole populations, where the relationship turns out to be much weaker than the individual rule implies. They are different questions and both answers are real.

The size of the effect is routinely overstated, and correcting it is in zinc's favour. In that same stratified work, zinc was among the less inflammation-sensitive micronutrients. The markers falling by more than 40% were selenium and vitamins A, B6, C and D. Zinc's threshold of 20 milligrams per litre of CRP was the most permissive of any micronutrient in that lab's dataset, against 10 for selenium and vitamins A and D and 5 for vitamins B6 and C. At population level, a large international project found the associations between zinc and inflammatory markers weak and inconsistent, and concluded there was insufficient evidence to adjust for it in women of reproductive age, though applying a correction elsewhere has moved national prevalence estimates by 10 percentage points or so, which is not nothing. What is not possible is a correction for an individual, because the change varies markedly from person to person.

Albumin, and why there is no corrected zinc

Around 60% to 70% of serum zinc is carried on albumin. A low albumin therefore lowers total zinc, and in a large population analysis a low albumin was the single largest factor associated with falling below the zinc cut-off, larger than pregnancy and far larger than anaemia.

No albumin adjustment for zinc has been validated for clinical use, and this is a genuine difference from adjusted calcium. The published advice is to read albumin alongside zinc qualitatively, where a low zinc with a normal albumin points one way and a low zinc with a low albumin and a raised CRP points somewhere else entirely. If you see an albumin-corrected zinc quoted anywhere, it is not supported.

The other things that move it

Pregnancy lowers serum zinc regardless of intake, by about 20% to 25% by the end, with the decline evident by 2 months and attributed later to plasma volume expansion.

The combined pill is worth a correction. Historic work on higher-dose formulations reported markedly lower serum zinc in users. On modern low-dose preparations the difference is small to absent, and a large recent survey found none. The large and undisputed oral contraceptive effect is on copper, not zinc, where the reference interval for women taking it is substantially higher.

Exercise moves it and the direction is genuinely disputed. Running to threshold and strength work to exhaustion have both been found to lower serum zinc significantly, and incremental cycling in trained endurance athletes has been found to raise it, tracking adrenaline and cortisol. Not testing straight after a hard session is reasonable advice. A stated direction and size is not supportable.

Corticosteroids are believed to lower it. The group that derived the cut-offs excluded people taking steroids when they did so, which tells you the effect was considered real, and we could not find a figure for it.

How much does a repeat have to move

Zinc is the opposite case to magnesium. Its index of individuality comes out at about 0.99, meaning within-person and between-person variation are similar, so the population reference range is more applicable to an individual for zinc than it is for magnesium or B12. Zinc's problem is not the comparator. It is everything to do with how and when the sample was taken.

More is not better

Zinc and copper compete for absorption, so sustained high-dose zinc supplementation can produce copper deficiency, which causes its own anaemia and, if it goes on long enough, nerve damage. The mechanism, which is more specific than simple competition, is in Minerals and bone, in the Body Guide. It is one of the most common self-inflicted deficiencies in people who supplement heavily.

The doses are not exotic. In controlled studies, 50 milligrams a day for 6 to 10 weeks measurably reduced the activity of a copper-dependent enzyme in red cells. Reviewing the published cases of zinc-induced blood problems, the doses ran from about 50 milligrams a day to over 1,500, the durations from weeks to years, and the sources included ordinary oral supplements and denture adhesive cream. Nearly all had anaemia, serum copper was reduced in every reported case in that series, and the bone marrow appearance led to frequent initial misdiagnosis as a marrow disorder. Stopping the zinc and replacing the copper corrected the blood picture in most within weeks to months. The neurological problems improved more slowly and were sometimes incomplete.

The European tolerable upper intake level for adults is 25 milligrams a day from all sources, set from a no-observed-adverse-effect level of 50 with the critical endpoint being copper status. Retail zinc tablets are commonly sold at 25 to 50 milligrams.

There is one important refinement to the advice to go and check your copper. Serum copper is a late marker. In those controlled studies the functional enzyme marker fell at 6 to 10 weeks while serum copper and caeruloplasmin stayed entirely normal. So a normal serum copper in someone taking 50 milligrams of zinc a day does not establish that nothing is happening. Neither test is one we sell, and both are ordinary tests a GP can arrange.

Units

UK labs report zinc in micromoles per litre. American reporting uses micrograms per decilitre. Multiply micromoles per litre by 6.54 to convert. So 10.0 is 65 micrograms per decilitre, 11.3 is 74, and 18.0 is 118.

Adjusted Calcium

Corrected calcium

Calcium in the blood, corrected for the amount of albumin present. Around half of circulating calcium is bound, mostly to albumin, and is not biologically active.

The adjustment exists for a real problem. Someone with low albumin will have a low total calcium while their active calcium is entirely normal, and the uncorrected number would be misleading. Where albumin is abnormal, the adjusted figure is the conventional way to read calcium.

The control system underneath this number, the 4 glands watching it, the receptor they watch it with, and the reason albumin means 3 different things on 3 different parts of a report are all in Minerals and bone, in the Body Guide. What follows is about the arithmetic, which is where the difficulty is.

In April 2026 three international bodies said to stop reporting this number

A joint position statement from the international federation of clinical chemistry's bone metabolism committee, the international osteoporosis foundation's working group and the European federation's chronic kidney disease committee recommends 3 things. Report total calcium as the default. Order a directly measured free calcium where calcium status is clinically important or hard to interpret. And use free calcium first-line where albumin is severely low or the person is on dialysis. Their stated reasoning is that the adjustment became embedded in practice despite limited validation against free calcium, and that the corrections can worsen diagnostic accuracy and mask clinically important abnormalities.

It is a short consensus piece and not a systematic review, and it does not say what labs without access to free calcium should do, or address primary care, which is where almost all calcium testing actually happens.

UK practice points the other way, and the disagreement is live. The UK clinical biochemistry body's standing position paper, from 2015, recommends that all labs providing total calcium should report adjusted calcium, using locally derived equations, with the measured total reported alongside it. NICE guideline NG132 instructs clinicians to measure albumin-adjusted calcium when testing for primary hyperparathyroidism and, in a separate recommendation, not to measure free calcium for that purpose. So NICE and the international bodies now point in opposite directions on the same question.

And this is not a disagreement that anyone has quietly let lie. In September 2026 a biochemistry department at an NHS trust in West Sussex published a reply to the April statement in the same journal. At the time of writing, UK lab practice has not changed and the 2015 paper stands. This is an argument in progress, and if your report carries an adjusted calcium it is because your lab is on the side of the argument that most UK labs are still on.

The formula is not universal, and the one everyone calls Payne is not Payne's

The original equation comes from 1973. It was derived from 200 consecutive specimens that happened to have been sent in for liver function tests, on one analyser, using one particular method of measuring albumin. It was checked against that same lab's own total calcium reference range, which is circular with respect to the thing it is meant to estimate. It was never validated against directly measured free calcium at all.

Its author said so at the time. The original paper states that the adjustment cannot be applied to data from another lab if the accuracy and precision of the calcium and albumin measurements, and therefore the normal ranges, differ greatly from theirs. That caveat has been ignored for 50 years.

And the version in widest UK use is not the original. The original regression slope converts to about 0.0247 millimoles per litre for each gram per litre of albumin. The formula most often printed in the UK uses 0.02, which is 19% smaller. At an albumin of 20 the 2 differ by about 0.09 millimoles per litre. Calling the 0.02 equation "the Payne equation" is common and it is wrong.

What UK labs actually use has changed. In 2008, of 101 labs responding to a survey, 58 used the original equation. By 2022, of 55 labs responding, only 9 did, and 42 of the 55 said their equation was locally derived. So the number of equations in UK use is now close to the number of labs, which is what the professional body asked for and which has a consequence.

How much does the formula change the answer

Four published formulae were applied to the same samples. The proportion of the same people they classified as hypercalcaemic ranged from 1.7% to 81%.

That is the number to remember. Same blood, same people, nothing changed except which equation the lab happened to use, and somewhere between 1 in 60 and 4 in 5 came out flagged.

The same effect shows up across real labs. Across 15 UK and Irish hospital labs, the prevalence of low calcium as judged by adjusted calcium ranged from 6% to 44% of inpatient samples. That range is driven by the lab, not by the patients, and it was higher in labs using the older albumin method.

The albumin method is most of the problem

There are 2 common ways of measuring albumin, and they do not agree. The older bromocresol green method reads higher than bromocresol purple, and the gap is not constant. At an albumin above 35 grams per litre it runs at about 3 grams per litre. Between 30 and 35 it is about 6. Between 20 and 29 it is about 7.5. Below 20 it reaches about 10. So it is worst exactly where the adjustment is most relied on.

The reason is that the older method is not specific to albumin. It also reacts with the globulins that rise during inflammation, which means the size of the error tracks how inflamed the person is. Measured directly on the same samples, adjusted calcium calculated from the 2 albumin methods differed by 0.12 millimoles per litre.

One caution on that, because it is usually told too simply. The older method is not a single method. Run with a short incubation it is much more albumin-specific and produces a constant offset instead of a widening one, and the person who derived the original calcium formula concluded on that basis that the newer method had no advantage over it. Papers reporting a widening bias and papers reporting a constant one are probably both right about their own configuration. The honest statement is a range of about 3 to 10 grams per litre, worse at low albumin and worse with inflammation, and which you got depends on your lab.

The adjustment makes agreement between labs worse, not better

This is the finding that ought to be better known. Using 6 months of UK quality assessment data grouped by platform, adjusted calcium showed a bias of 11%, and the scatter of adjusted calcium results exceeded the scatter of plain total calcium results even after excluding the worst-performing platform. Importing albumin's error into the calcium result does not cancel anything. It adds.

An 11% bias is 0.24 to 0.29 millimoles per litre across the reference interval. For scale, the harmonised UK reference interval for calcium is 0.40 millimoles per litre wide, so between-lab variation in adjusted calcium is roughly 60% to 70% of the entire normal range.

Set that against the thresholds it is used with. NICE's 2 action figures for adjusted calcium are 2.5 and 2.6 millimoles per litre, which are 0.10 apart. The between-lab uncertainty in adjusted calcium is roughly 2.5 times the gap between them. And the same 11% is about 6 times the amount a person's own calcium varies from week to week. Which lab ran the sample matters several times more than what your own calcium actually did.

A 2026 UK quality assessment study puts a number on the consequence. Testing total calcium against reference-method values, no commercially available analytical approach achieved adequate performance on 3 specimens, with platform biases running from about 3% low to about 12% high against an allowable figure of roughly 1%. On one specimen, 16 of the 91 UK participants classified a sample that was hypercalcaemic as normal.

How well the adjustment actually performs, and the direction it fails in

Measured against directly measured free calcium, which is the real reference, adjusted calcium generally performs no better than plain uncorrected total calcium and in several comparisons worse. In a study of over 20,000 patients the agreement was significantly less than or equal to that of total calcium. In a separate study predicting free calcium, the adjusted figure was frankly worse than the unadjusted one. In 250 intensive care patients, no established correction formula outperformed the plain total calcium. In intensive care and in dialysis it performs badly enough that the investigators recommended abandoning it in those settings.

The direction of failure is the part that is usually stated backwards, and it is stated backwards almost everywhere. It depends entirely on which side of an albumin of 40 grams per litre the person sits, because that is the figure the formula normalises to.

Below 40, the correction adds to the calcium, so the adjusted figure comes out higher than the measured one. A genuinely low free calcium gets adjusted up into the normal range. Low calcium is therefore masked, and the size of this is not small. In studies of sick populations, more than half of true low calcium results were reported as normal, and in one intensive care series 85% were.

Above 40, the correction subtracts, so the adjusted figure comes out lower than the measured one. A genuinely raised calcium gets adjusted down towards normal. High calcium is therefore masked. Measured against directly measured free calcium in one series, the underestimate grew steadily above an albumin of 40 and reached about 0.20 millimoles per litre above 44, which is larger than the correction term itself because the underlying relationship is not the straight line the formula assumes. Half the cases of genuinely raised free calcium in that series were missed this way.

And that second direction is the one that applies to most people buying a blood test. A fit, well-hydrated adult with a high-normal albumin of 45 to 50 has their calcium adjusted downward. The failure mode people worry about, being told their calcium is low when it is not, is the hospital one. The failure mode that applies here is the opposite.

The counterweight, which is real and which we are not going to leave out

Almost the entire case against adjusted calcium comes from hospital populations, which is to say intensive care, dialysis, tertiary inpatients and people with severely low albumin. Our customers are ambulant, self-selected and mostly have a perfectly normal albumin, and in that population the picture is materially better.

In ambulant patients after thyroid and parathyroid surgery, venous adjusted calcium and directly measured free calcium were equivalent, with a mean difference of 0.03 millimoles per litre. Where the adjustment failed in the large studies, the failure was specific to abnormal albumin, reduced kidney function or a disturbed acid-base state. And the setting-matching matters as much as the equation. An equation derived from a community population correctly identified calcium status in 92% of healthy participants, where an equation derived from inpatients managed 46% on the same people.

So the honest summary is that the adjustment is roughly harmless and mildly helpful in a healthy ambulant adult with a normal albumin, and becomes actively misleading when albumin is low, when kidney function is reduced, or when someone is acutely unwell. Which is to say it fails in exactly the people a test bought online is least able to interpret.

The tourniquet moves both numbers, and by different amounts

One minute of venous stasis raises albumin by about 3.5% and calcium by about 1.6%. Three minutes raises albumin by about 8.6% and calcium by about 3.6%. The European recommendation for venous sampling is a total tourniquet time of up to 1 minute, and measurable changes in albumin and calcium appear at 60 seconds.

Here the adjustment earns its keep, and the arithmetic is worth doing because it is easy to get backwards. Because the formula subtracts albumin, and albumin rises faster than calcium under stasis, the adjustment cancels most of the artefact. Starting from an albumin of 40 and a calcium of 2.40, a 3-minute tourniquet raises the measured total calcium by about 0.09 millimoles per litre and the adjusted calcium by under 0.02. Both still go up. The adjustment would have to be fed an albumin rising more than 3 times as fast as calcium to push the adjusted figure down, and under stasis albumin rises about twice as fast. So a long tourniquet inflates both numbers and inflates the total far more, which is one of the better arguments for reporting the adjusted figure at all.

Posture does something similar and smaller. Standing rather than lying shifts plasma volume by a substantial amount, which raises albumin and everything bound to it. European guidance asks for 15 minutes seated before the sample.

About fasting, and a correction to what we used to say here

It is widely repeated, including by us, that calcium rises after a meal. We went looking for the evidence and could not find it. European guidance lists calcium among analytes affected within 4 hours of eating but publishes no figure, and the one controlled study we could find, sampling volunteers at 4 and 6 hours after a meal, found no change in any analyte exceeding the reference change value. We are not aware of published evidence that an ordinary meal meaningfully shifts calcium or adjusted calcium at those intervals, though the first hours after eating have been less well studied.

The generic recommendation for a morning fasted sample stands on its own merits and applies to the whole panel. It is not specifically about calcium, and we should not have implied that it was.

The wrong tube, and how often it explains a low result

EDTA from a full blood count tube chelates calcium, so carryover lowers it, and how much depends on the method. On one platform at around 3% contamination by volume, calcium fell by only about 1% while potassium rose by 12% and zinc fell by 71%, which is a long way from the dramatic calcium collapse the folklore describes. Gross contamination is a different matter and does lower calcium substantially.

The useful figure runs the other way round. In one UK lab over a month, 14.3% of markedly low adjusted calcium results, meaning below 2.0 millimoles per litre, turned out to be EDTA contamination and not the patient. Across all contaminated samples, 92% came back normal after being re-bled. So roughly 1 in 7 strikingly low adjusted calcium results in that series was the tube.

Two things that do not go wrong

Haemolysis is not much of a calcium problem. On commonly used platforms calcium is unaffected up to a haemolysis index of around 800 and albumin up to around 1000, with the exact limits set by the manufacturer, and neither featured among the analytes meaningfully affected in 2 recent interference studies. A haemolysed sample is a reason to doubt the potassium.

Stability is the other, and it is the reason this test can be sold by post at all. In a study of 81 analytes, most were stable for 24 hours under all storage conditions before centrifugation, and the ones that were not included potassium, phosphate, magnesium, glucose and parathyroid hormone. Neither calcium nor albumin appears in that list. The awkward corollary is that parathyroid hormone does, and parathyroid hormone is exactly what NICE asks for next after a raised calcium, which is one of several reasons it is not a test we offer.

How much does a repeat have to move, and a claim we are not going to make

Calcium is famously tightly regulated, and the tempting conclusion is that it has a very low index of individuality, meaning the population range fits an individual badly. We checked, and it does not. Published estimates put calcium's index of individuality at around 0.8, which is intermediate to high, so the population range is reasonably applicable to an individual for this marker. Reported values do vary between studies. Anyone telling you calcium has marked individuality is stating something checkable and probably wrong, and we are not going to repeat it.

The real argument is different and it is stronger. The reference change value for total calcium is about 5.6%, which is about 0.13 millimoles per litre at a typical level. The harmonised reference interval is 0.40 wide. So the whole normal range is only about 3 significant steps wide. A person can move from the bottom of normal to the top of it in 3 statistically real steps and never once be flagged as abnormal. That is the case for comparing yourself with yourself, and it does not depend on the individuality argument at all.

There is a second consequence, and it is slightly uncomfortable. Because the adjustment imports albumin's variation into the calcium result, it raises the reference change value from about 0.13 to about 0.15. In a stable, healthy person whose albumin barely moves, the adjustment adds noise without adding information. In a sick inpatient whose albumin genuinely swings, the opposite is true, and the original work showed adjustment reducing within-person variation substantially in hospital patients. Both are right, about different people.

What the guideline does with an abnormal result

This is what NICE directs clinicians to do, and it is set out here because the structure of it is the point, not because this page can tell you anything about your own number.

NG132 asks for an adjusted calcium in people with symptoms of a raised calcium such as thirst, frequent urination or constipation, in osteoporosis or a previous fragility fracture, with a renal stone, or where a raised adjusted calcium of 2.6 or above has turned up incidentally. It then says to repeat the measurement at least once if the first is 2.6 or above, or 2.5 or above with features of primary hyperparathyroidism. Parathyroid hormone is measured only when the adjusted calcium is 2.6 or above on at least 2 separate occasions, or 2.5 or above on at least 2 occasions with suspicion. And the interpretation is against the midpoint of the local lab's own parathyroid hormone reference range, because NICE declines to publish an absolute figure for it.

So the guideline is built on 2 separate occasions, a lab-specific comparator, and a clinician weighing 2 numbers against each other. By its own construction, a single adjusted calcium is not an answer.

What a persistently raised adjusted calcium turns out to be depends enormously on where the blood was taken. The claim that primary hyperparathyroidism and cancer between them account for the great majority of raised calcium appears everywhere, and it is a hospital statistic. In one hospital series cancer was much the largest cause and primary hyperparathyroidism a small minority. In primary care followed over 10 years, primary hyperparathyroidism was the largest single cause and cancer a much smaller fraction. Same diseases, opposite proportions, entirely because of where the sample came from. Anyone quoting the hospital pattern to someone who bought a test online is quoting the wrong population.

Free calcium, and why it is a hospital test

Directly measured free calcium is the definitive measurement and the reference every study above is measured against. The reason it cannot be sold by post is not squeamishness, and it is specific.

Free calcium depends on pH. Every 0.1 fall in pH raises it by about 0.05 millimoles per litre, because hydrogen ions compete with calcium for the same binding sites on protein. Carbon dioxide escaping from an opened or partly filled tube raises the pH and therefore lowers the measured result, so collection has to be anaerobic and the caps have to stay on. If analysis will be delayed beyond 30 minutes the sample has to go into an ice slurry, and even then whole blood keeps for about 4 hours. Ordinary lithium heparin binds calcium and falsely lowers the result, so the anticoagulant has to be calcium-balanced and correctly concentrated, and the fill volume has to be right. None of that survives an overnight post.

Handheld devices have been evaluated for exactly this purpose. The most promising one on the market was accurate in standardised solutions, ran about 10% out against a blood gas analyser on venous blood and about 18% out on capillary blood, and the investigators concluded it was not yet suitable for capillary or home testing. The failure mechanism is the one above, which is carbon dioxide escaping from an open system.

And there is an honest finding in that study which cuts in the adjustment's favour. In those same ambulant patients, venous adjusted calcium agreed with the blood gas analyser's free calcium better than the handheld free calcium device did. A well-handled adjusted calcium beat a badly-handled direct measurement.

Pregnancy is the case the adjustment was built for

Albumin falls through pregnancy because of haemodilution and stays low until delivery. Total calcium falls with it, to values that can sit well below the reference range, while free calcium stays constant. So the fall in total calcium in pregnancy is an artefact of the albumin, which is precisely the situation the adjustment was designed to correct, and adjusted calcium remains normal.

This is the clearest counter-case to the blanket 2026 recommendation, and there is no widely adopted pregnancy-specific reference interval for adjusted calcium to read it against.

Drugs

Thiazide diuretics raise calcium mainly by increasing how much the kidney reabsorbs, and the clinically important framing is that they unmask an existing problem instead of causing one. A thiazide-associated raised calcium is frequently an underlying parathyroid problem that was previously intermittent or silent, which is a good reason not to write a raised calcium off as the diuretic.

Lithium raises calcium through a drug-induced overactivity of the parathyroid glands. Pooled across the published work, hypercalcaemia appears in about 3% of people taking it by total calcium and about 4% by free calcium, and hyperparathyroidism runs at around 4% against roughly 0.5% in the background population. The authors of that work note that calcium is insufficiently measured in people on lithium in practice.

High-dose vitamin D raises calcium, and the figures for how much and at what dose are set out under Vitamin D.

On calcium supplements, a common question is how long before a test to stop them. We looked specifically and found no primary study measuring how much, or for how long, an oral calcium dose raises serum calcium in a healthy adult. We are not going to make a figure up.

So how should the number be read

As an estimate with known limits, calculated with your lab's own equation from your lab's own albumin method, varying between labs by considerably more than the thresholds it is compared against, adjusted downward if your albumin is high and upward if it is low, and currently the subject of an unresolved disagreement between UK practice and 3 international bodies.

It is still what every UK lab reports and it is still the conventional way to read calcium where albumin is abnormal. It is also worth reading the albumin on its own account instead of trusting the arithmetic to have dealt with it, and worth remembering that the hormone actually running calcium is not on this panel at all.