Full Blood Count Markers

Haemoglobin

Hb

The oxygen-carrying protein inside red blood cells, and the number most people mean when they say anaemia. It measures the concentration of haemoglobin in blood and not the number of cells. Each molecule is built around iron, which is why the 2 marker groups are linked, and iron itself is in Iron, in the Body Guide.

Dehydration concentrates it and over-hydration dilutes it, so it moves a little with fluid status independently of anything else. The size of it is easy to overstate, so here it is. Cycling for 2 hours in the heat drops plasma volume by about 2% if you drink water and about 4% if you drink nothing, and haemoglobin, haematocrit and red cell count all come out highest when unhydrated. A full day of fluid restriction followed by resistance exercise does far more, with plasma volume down 12.5% 5 minutes after the session.

One thing runs against intuition. Prolonged endurance events do not reliably concentrate blood. In Ironman triathletes who lost a median 2.5 kilograms of body mass, plasma volume increased by a median 10.8%. So a hard day in the heat concentrates these markers, and a very long event may dilute them.

Low haemoglobin is anaemia, the most common causes being iron, B12 or folate deficiency, blood loss, and chronic disease. High haemoglobin carries the same questions as a high haematocrit.

Where the numbers come from

The World Health Organization's definition of anaemia is the committed number here, and it is one of the most widely used thresholds in medicine: haemoglobin below 130 grams per litre in men and below 120 grams per litre in non-pregnant women.

WHO reissued that guideline in March 2024 and kept both adult figures unchanged. What changed sits outside them. The threshold for children aged 6 to 23 months came down from 110 to 105, pregnancy became trimester-specific with the second trimester at 105, and the altitude correction was rebuilt from the ground up.

The men's figure is weaker than it looks, and WHO says so

The work WHO commissioned to set these thresholds excluded anyone with a condition that could lower haemoglobin, then took the fifth centile of what was left.

For non-pregnant women aged 18 to 65 it came out at 119.7 grams per litre, which lands on 120 almost exactly. For men aged 18 to 65 it came out at 134.9, not 130. The guideline group looked at that, considered moving the men's threshold to 135, judged the evidence uncertain and kept 130. Their own words were that data in adult men may suggest a higher cutoff but that there was uncertainty in the evidence.

So the women's threshold is empirically confirmed and the men's is a deliberately retained convention sitting about 5 grams per litre below where the healthy-population data fall. So a man at 132 is above the anaemia threshold and below the healthy fifth centile at the same time.

The same analysis found no evidence that these thresholds should differ by ethnicity, which is the counterpart to the neutrophil entry further down, where a genetic variant does shift a reference range and ethnicity turns out to be a poor way of predicting it.

Altitude and smoking, with the numbers

These are the 2 exposures WHO thinks large enough to publish formal corrections for.

Altitude is much the larger. The 2024 correction starts at 500 metres, down from 1,000 in the old version, and runs as a continuous equation instead of bands, reaching about 33 grams per litre at 4,500 metres.

That change was not cosmetic. Recalculating global anaemia under the new elevation adjustment raised the 2023 worldwide prevalence from 24.0% to 26.4%, which is 198 million people newly counted as anaemic, and moved anaemia from the third to the second largest cause of disability globally. The largest increases were between 500 and 2,000 metres, which is exactly the band the old version ignored.

Smoking is much smaller than its reputation. In the data behind the 2024 revision, smokers averaged 3.3 grams per litre higher than non-smokers, falling to 2.3 once altitude was accounted for. The older corrections under-adjusted light smokers and low altitudes and over-adjusted above 3,000 metres. The 2 corrections add together.

Anaemia is a finding and not a diagnosis. The threshold tells you the oxygen-carrying capacity is low. It does not tell you why. The red cell indices below, read alongside the iron markers and B12, are what separate the causes.

Sickle cell trait does not show up on a blood count

Sickle cell trait is often described as causing a mild anaemia. It does not. The blood count of a carrier is normal, with a normal haemoglobin, a normal average cell size, a normal MCH and a normal red cell count. Carriers are healthy and unaware of their status unless they are screened, and nothing on this panel will show it. Haemoglobin separation or a solubility test is what finds it.

So a low haemoglobin or a small average cell size in someone who carries sickle cell trait has a different cause, most often low iron or a co-inherited alpha thalassaemia, and it needs working out on its own terms rather than being put down to the trait.

The national screening programme gives carriers a single caution, which is to take care where the oxygen supply may drop, for example under general anaesthetic, diving or climbing. It also states plainly that carriers do not have a condition requiring treatment.

G6PD deficiency is common and leaves nothing behind between episodes

Glucose-6-phosphate dehydrogenase deficiency destroys red cells only when something oxidising arrives, usually an infection, broad beans or a short list of drugs. Between episodes there is nothing to see. Haemoglobin, bilirubin, LDH, haptoglobin, the reticulocyte count and the blood film are all normal, and no routine panel has a marker that moves. The common forms do not cause ongoing red cell destruction.

It is not rare here, and it is most common in men of Black, South Asian, Middle Eastern and Mediterranean background. How common it is, and what it does to an HbA1c, is in HbA1c, in the Biomarker Guide, where it matters more.

How an episode of red cell destruction reads on this panel

Haemoglobin falls, sometimes over hours. The reticulocyte count rises as the marrow replaces what was lost, peaking around 5 to 7 days after the trigger, and because reticulocytes are large that pulls the average cell size up. Bite cells and blister cells appear on the film. Off this panel, LDH rises, haptoglobin falls as it binds free haemoglobin and is cleared, and bilirubin rises in its unconjugated form while ALT and ALP stay where they were. A raised bilirubin alongside normal liver enzymes is a red cell pattern and not a liver pattern.

How this whole panel behaves in the post

Haemoglobin is the most robust number on the blood count, so it is the reference point for everything that follows about the ones that are not.

A blood count is not a measurement of dissolved substances. It is a count and a sizing of living cells, and cells keep changing after the blood leaves the arm. Red cells take up water and swell. White cells begin to break down, and the 5 types do so at different rates. Platelets swell and stick together. None of that stops because the tube is in a padded envelope.

Our own lab publishes the window. Inuvi confirm the blood count as stable for up to 72 hours at 4 degrees and at room temperature, and say that outside those temperatures it degrades faster, especially the white cell count and the differential. They also name the parameters that a delay affects, which are MCV, MCHC, red cell distribution width, haematocrit and the white cell differential. Around 98% of samples reach them inside 3 days, so their validated window and their normal transit time are the same number. That is why we ask for collection Monday to Thursday and posting the same day.

The published research behind that runs out to 48 hours at 4, 10 and 23 degrees. At room temperature, as a percentage of the starting value at 24 hours.

Haemoglobin, red cell count and MCH were unchanged, drifting by 1.1% or less all the way to 48 hours. Haematocrit had risen, MCV was up 4.5%, red cell distribution width up 8.4% and MCHC down 4.7%. The total white cell count still looked normal. Underneath it, neutrophils had fallen to 47% of where they started, eosinophils to 47%, while lymphocytes had risen to 172% and basophils to 262%. At 48 hours neutrophils were at 24%.

Warmth makes it worse. In a separate study at 33 degrees, closer to a parcel in a car in July, MCV rose 10.6% by 24 hours and 16.7% by 72, taking an average cell size of 87.8 femtolitres to 101.8. Haematocrit rose 17.6% over the same period. Refrigeration protects most of the red cell numbers well and does not protect the platelet size at all.

Three things about that matter more than the numbers themselves.

The lab flags the clock, not the drift. A sample that arrives outside its stability window is reported as DELAY, and one judged to have been affected by postal conditions is reported as AGED, so a late sample does not pass silently. What no flag can do is measure the drift itself. An MCV that has risen 4.5% and an MCHC that has fallen 4.7% are arithmetically consistent with each other, sit inside the reference range, and look exactly like a real result from a real person. The analyser cannot detect red cell swelling. Internal quality control and external quality assessment both run stabilised material, so they certify the instrument and say nothing about your sample. The one check that would catch it, comparing against a previous result from the same person, needs a previous result.

The drift is directional, and it points at disease. These are not random errors that average out. Delay pushes the average cell size up, the distribution width up, the MCHC down, the haematocrit up, the neutrophil count down and the platelet count down. Every one of those is a movement towards something that looks abnormal.

The parts that fail first are not the parts you would guess. Average platelet size goes first, breaching acceptable limits inside 4 hours when it is warm and drifting at every temperature by about a day. Blood film morphology can become compromised as early as 4 hours at room temperature. A manual differential, eosinophils and other white cells are compromised after about 8-12 hours. Haemoglobin, red cell count and MCH are the last to go, and on that evidence they are usually still fine at 48 hours.

Each entry below says what postal delay does to that specific number and in which direction, because that is more useful than a general warning. Where a marker is not dependable on a posted sample, it says so.

Units

The UK reports haemoglobin in grams per litre, the US in grams per decilitre. Divide by 10, so 130 grams per litre is 13.0 grams per decilitre.

In the post

The most robust parameter on the panel, and unchanged at 48 hours at every temperature tested. It only fails when a sample gets genuinely hot, at around 37 degrees.

Haematocrit

Hct, packed cell volume

The proportion of blood volume made up of red cells, expressed as a percentage. It tracks closely with haemoglobin.

It rises in dehydration, at altitude, in smokers, and on testosterone therapy.

A persistently high haematocrit makes blood more viscous, and that is the mechanism behind the concern about clots, stroke and heart attack. It is why haematocrit is monitored on testosterone replacement, where a rise is common. How much of that risk is caused by the haematocrit itself, and how much simply travels alongside it, is a good deal less settled than the monitoring implies, and the evidence is set out below. A low haematocrit is the picture of anaemia. Persistently high readings can also raise the question of polycythaemia vera, a bone marrow disorder in which red cells are overproduced, though most high readings have more ordinary explanations.

Where the numbers come from

The committed numbers here are all about testosterone, and they do not agree with each other. The spread is wider than usually reported.

In the UK, the British Society for Sexual Medicine guideline, published in 2023, says haematocrit should remain below 54%, with dose reduction or a change of preparation above 0.54, stopping and reintroducing at a lower dose if it stays high, and dose adjustment or periodic venesection where needed.

The European Academy of Andrology, in a 2020 guideline endorsed by the European Society of Endocrinology, is the most conservative of the set and is usually quoted as though it were the most permissive. Its published position is that a haematocrit above 48 to 50% is a relative contraindication to treatment.

In the US the Endocrine Society, still on its 2018 guideline, advises withholding testosterone where haematocrit exceeds 54% until it normalises, then resuming at a lower dose, and lists therapeutic phlebotomy as an effective option. The American Urological Association, published in 2018 and formally revalidated in 2024, uses 50% as the point to ensure a pre-treatment level is below before starting and to investigate a cause, and 54% as the point to reduce the dose or stop.

So the real range across bodies is 48% to 54%, and not 50% to 54%.

What is actually known about the risk

Said plainly, because 54 gets quoted as though it were settled.

Reviewed in 2025, the evidence supporting any specific haematocrit threshold is lacking, and considerable uncertainty remains about how best to manage it. Men on testosterone who develop erythrocytosis do have a higher rate of thromboembolic events, so the finding is not that the risk is imaginary. It is that the particular number is a consensus landing point and not a measured cliff edge. In practice, blood is being taken off anywhere between 50% and 52%, which is below every guideline threshold above.

The one place a haematocrit target has been tested in a randomised trial is polycythaemia vera, and there it worked. In people with the JAK2 mutation, holding haematocrit below 45% instead of between 45 and 50 cut cardiovascular death and major thrombosis from 9.8% to 2.7% over about 2 and a half years, which is close to a 4-fold difference. Note both boundaries of that finding. The threshold that mattered was 45%, which is 9 points below the testosterone figures, and the people in it had a bone marrow disorder.

In the testosterone setting there is still no randomised evidence for any threshold. The largest cardiovascular safety trial found testosterone no worse than placebo for major cardiovascular events. Erythrocytosis was the most reproducible adverse effect but it did not translate into major cardiovascular events. Pulmonary embolism and atrial fibrillation however, did show some statistically significant increases.

The work most often cited on the other side is a database analysis, and it found that a haematocrit at or above 52% came with more major cardiovascular events and venous clots, 5.15% against 3.87%. Its own opening sentence is that an unsafe haematocrit threshold for men receiving testosterone has never been tested.

And then there is the timing problem, which is the strongest argument against a direct causal link. In every testosterone study that reported it, the people who clotted were not erythrocytotic when it happened. In one group of men on testosterone, 10 developed a deep vein thrombosis, 7 of them with another identifiable risk factor such as recent surgery, long-haul travel or an inherited clotting disorder, and not one of the 10 had erythrocytosis at the time of diagnosis. Among transgender men, 5 had a thromboembolic event, only 4 had ever met the criteria for erythrocytosis, and again none had it at the time.

Haematology's own position is blunter than the andrology guidelines. A 2021 review states that management of erythrocytosis outside polycythaemia vera has been confounded by unfounded concerns about thrombosis risk, and that venesection is seldom warranted, with frequency better determined by symptoms than by a haematocrit threshold. A 2025 teaching review from the American Society of Hematology says there are currently no evidence-based treatment guidelines for erythrocytosis without the JAK2 mutation, and notes that it is conventionally defined using the same haemoglobin and haematocrit thresholds as polycythaemia vera, which is to say the thresholds are inherited convention and not independently derived.

How it is produced, and how it is reported

Whether haematocrit is measured or calculated depends on the analyser. On most instruments the average cell size is the measurement, with the machine recording the volume of each red cell as it passes and taking the mean, and haematocrit is then computed by multiplying that average by the count. Some instruments work the other way round, adding up the volumes to get haematocrit directly and dividing by the count to get average size. Both are correct, and the first arrangement is much more common.

What is always true on an automated analyser is that haematocrit and average cell size are never both measurements. One is derived from the other. The older method of spinning a tube and reading off the packed red cell column is a genuinely separate physical measurement, and it reads slightly higher, because some plasma stays trapped between the packed cells.

UK labs often report it as a fraction instead of a percentage, so 0.54 instead of 54%. The same number, the decimal point moved.

In the post

Rises with delay, because the red cells swell, by about 8% at 48 hours at room temperature and more in the heat. That matters more here than anywhere else on the panel, because the threshold in question is 54% and the drift runs towards it.

Red blood cell count

RBC

The number of red cells per unit volume. Read together with haemoglobin and haematocrit, it helps distinguish having fewer cells from having cells that each carry less haemoglobin.

Where the numbers come from

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

It is one of the numbers that is always genuinely measured, along with haemoglobin and the white cell and platelet counts. Of haematocrit and average cell size, one is measured and the other derived depending on the analyser, and MCH and MCHC are always arithmetic, so most of the rest of the red cell panel is built from the count and the haemoglobin.

Variation inside the reference range has not been shown to predict anything actionable for the red cell count, or for haemoglobin and haematocrit. The associations in population data are small.

In the post

Robust, and unchanged at 48 hours at every temperature tested. Among the last parameters to fail.

MCV

Mean corpuscular volume

The average size of a red blood cell, and one of the most useful single clues in a blood count.

Small cells point in a different direction from large ones. Iron deficiency and thalassaemia trait produce smaller cells. B12 or folate deficiency and alcohol produce larger ones, as do some medications. An underactive thyroid appears on most lists of causes of a raised average cell size. It is widely repeated, not well supported by primary research, and best treated as a possibility and not an established cause.

So the average cell size often tells you where to look next before anything else does.

Where the numbers come from

There is no published healthy threshold for this marker. The population reference range is assay-dependent, and a commonly quoted one is 77 to 95 femtolitres.

Notably, the UK's national screening programme for thalassaemia and sickle cell uses average cell size descriptively and puts its decision threshold on MCH instead. That is set out in the MCH entry below, and the reason turns out to be partly about how the 2 numbers survive being transported.

A normal average cell size does not exclude iron deficiency

This is the most common way this marker misleads, and the honest version is stronger than the usual phrasing.

Red cell size is one of the last things to change as iron stores fall. Compared against bone marrow iron staining, which is the reference standard, serum ferritin is by far the most powerful test and no red cell index comes close to it. In women of reproductive age, criteria built around average cell size found iron deficiency in 13.4% against 17.1% for criteria built around ferritin, so the cell size misses roughly a fifth of it. How iron is stored and moved is in Iron, in the Body Guide.

What does not exist, and this is itself a finding, is a clean published sensitivity for average cell size alone against ferritin or marrow iron in a general adult population. The claim is directionally right, near-universally believed, and quantitatively unanchored. Ferritin is the marker that answers this question.

The reverse also holds, and gets said less. A low average cell size is not specific to low iron stores. Normal marrow iron has been documented alongside a low average cell size and a low MCHC in the same people.

Telling iron deficiency from thalassaemia trait

Both produce small cells, and a long list of formulas has been published to separate them from the blood count alone, the best known dividing average cell size by red cell count. None of them performs well enough to rely on, which is why the UK pathway uses a haemoglobin electrophoresis test instead of a formula.

The recent work adds 2 things. A 2024 classifier picks out thalassaemia trait correctly around 9 times in 10, and its categories explicitly include thalassaemia trait with a normal average cell size as a real and separate class. And 2026 work across 4 different analysers found that 4 numbers, MCH, MCV, haemoglobin and red cell distribution width, are enough to sort the basic types of small-cell anaemia at better than 80% accuracy, with MCH the most informative of the 4. Adding more numbers does not push it past 90%.

Iron does not fix the small cells of thalassaemia trait

The NHS sickle cell and thalassaemia screening programme says this in a single sentence, that iron supplements must not be given to correct the apparent anaemia caused by carrier status, as they are ineffective. It also notes that carriers may be misdiagnosed as iron deficient, because their red cells look abnormal. NHS patient information adds that the mild anaemia some carriers have is different from iron deficiency anaemia and needs no treatment.

The reason usually attached to that advice is not supported. Iron supplements are widely said to cause iron overload in people with thalassaemia trait. That happens in thalassaemia as a disease, where transfusion and ineffective red cell production drive it, and the carrier state is not that. The sound reasons not to take it are that it does not work, it has its own side effects, and it delays finding out what is really going on.

A more damaging mistake runs the other way. Carrier status is no protection against iron deficiency, and the 2 coexist often, in exactly the groups where carrier status is common. Treating a low MCH as explained because a parent has thalassaemia trait misses ordinary iron deficiency that would have responded to treatment. Ferritin answers that question and no red cell index does.

Iron deficiency is sometimes said to drag haemoglobin A2 down far enough to hide beta thalassaemia trait, so the screen has to wait. Better designed work looking specifically at carriers found no meaningful effect, and the NHS antenatal algorithm measures it on the booking sample whatever the iron status.

In the post

The most replicated failure on the panel. Red cells absorb water and swell, so the measured size climbs, by 4.5% at 24 hours at room temperature and 10.6% in the heat. The consequence is that delay reports small cells as normal and normal cells as large, so a summer postal sample can manufacture a macrocytosis in someone whose red cells are an ordinary size.

MCH

Mean corpuscular haemoglobin

The average amount of haemoglobin inside each red cell. It usually moves in step with average cell size and is interpreted alongside it.

Where the numbers come from

There is a published UK threshold for this marker, which makes it unusual on this panel, and it is a national screening threshold and not a research figure.

The British Society for Haematology's 2023 guideline on screening for significant haemoglobinopathies puts the decision on MCH. An MCH below 27 picograms alongside a raised haemoglobin A2 is what indicates beta thalassaemia trait, and average cell size sits beside it descriptively. So MCH carries the decision and average cell size is descriptive. That is current UK practice and almost nobody quotes it. The NHS screening programme handbook that carried the same 27 picogram threshold was withdrawn in 2021, so the guideline is now the live source for it.

The guideline gives the reason in one line. MCH is used to screen for thalassaemia because it is a more stable parameter than average cell size.

It is haemoglobin divided by the red cell count, so both of its inputs are directly measured and neither is derived from anything else. That makes it one of the more dependable of the red cell indices, and the 2026 classification work mentioned in the average cell size entry found it the single most informative parameter for sorting out the causes of small red cells.

A low MCH with a normal haemoglobin A2 is where the pathway stops

The threshold above works in the direction it was written for. An MCH below 27 picograms with a raised haemoglobin A2 indicates beta thalassaemia trait. It does not work in reverse. A normal haemoglobin A2 speaks to beta thalassaemia trait and to nothing else, and the usual remaining explanation for a persistently low MCH in someone with normal iron stores is alpha thalassaemia trait.

Alpha thalassaemia carrier status cannot be confirmed without DNA analysis. That is the screening programme's own wording, and it describes a real dead end. The chemistry used for the NHS screen, and by every private lab, measures haemoglobin A2 and picks up structural variants, and it says nothing about the alpha genes. So a person can have a low average cell size, a low MCH, a normal haemoglobin A2, a normal haemoglobin separation and no answer, indefinitely. The programme is not looking for it either, since it targets the alpha zero form, most common in East and Southeast Asian and Mediterranean family origins, and not the alpha plus form usually seen in African, Caribbean, South Asian and Middle Eastern populations.

Haemoglobin E trait, which matters more for people of Bangladeshi background

Haemoglobin E is a variant haemoglobin that also produces a low average cell size and a low MCH, usually without anaemia, and it does show on a haemoglobin separation. UK consumer health writing almost never mentions it, and for people of Bangladeshi background it is the larger of the 2 stories.

The national carrier survey in Bangladesh put haemoglobin E trait at 8.7% and beta thalassaemia trait at 2.2%. Read those as figures for Bangladesh. No UK study has measured carrier frequency in British Bangladeshi, Pakistani or Indian populations, so what exists is country of origin data attached to ethnicity rather than anything measured here.

What the NHS screening pathway actually covers

Antenatal screening for thalassaemia is universal in England, because red cell indices are run on every booking sample, and that is why this threshold exists at all. Screening for sickle cell and other haemoglobin variants is not universal. Everyone is offered it in high prevalence trusts, and elsewhere the Family Origin Questionnaire decides. That questionnaire is completed for both biological parents in every pregnancy wherever you live, because it is also used to read the red cell indices. Newborn screening does not fill the gap, because the day 5 blood spot looks for sickle cell disease and cannot identify beta thalassaemia carriers.

In the post

It barely moves, drifting by 1.1% or less over 48 hours, and by 0.22% in the heat where average cell size moved 10.6%. It inherits the stability of its 2 inputs, which is why it is the more trustworthy of the size-related indices on a posted sample and why the UK screening pathway puts its threshold here.

MCHC

Mean corpuscular haemoglobin concentration

The average concentration of haemoglobin within a given volume of red cells, as opposed to per cell.

It is haemoglobin divided by haematocrit, so it is always arithmetic and never a measurement.

It moves in both directions for different reasons, and that is the whole entry

MCHC is the parameter labs lean on as a check that a sample is intact, because it is sensitive to 2 opposite kinds of damage.

Haemolysis pushes it up. If red cells rupture, haemoglobin is released into the plasma while the haematocrit falls, and the arithmetic produces an implausibly high value. That is what makes it a useful flag, and it is why a badly damaged sample gets rejected.

Delay pushes it down, and this is the more common problem for a posted sample. Cells swell, the haematocrit rises, the haemoglobin stays where it was, and dividing a flat number by a rising one gives a falling result. It was down 4.7% by 24 hours and 8.4% by 48 at room temperature, and down 9.4% by 24 hours and 12.5% by 72 at 33 degrees. Refrigeration protects it, out to 168 hours in one study.

A few other things produce a spuriously high reading without any haemolysis. Cold agglutinins, which make red cells clump in the tube, will do it, as will a very fatty sample, a very high bilirubin, and a very high white cell count.

Where the numbers come from

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

A raised value is usually not a red cell problem at all. Where multiple series of samples with a high MCHC were worked through, most were just red cell clumping, an optical interference or no identifiable cause, and only about 1 in 7 were an actual red cell disease such as hereditary spherocytosis, sickle cell disease or HbSC disease. A high value is a reason to look at the sample before it is a reason to look at the person. A low one usually travels with a low average cell size instead of saying anything on its own.

In the post

Falls, for the arithmetic reason above, and among the earlier parameters to go at room temperature. Well protected by refrigeration.

RDW

Red cell distribution width

How much the size of red cells varies from one to the next. A high value means a mixed population of large and small cells.

It is useful early. It often rises before average cell size moves out of range, and it helps separate a developing deficiency from a stable long-standing pattern.

Where the numbers come from

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

One measurement point trips people up when comparing 2 reports. There are 2 versions of this number, one expressed as a coefficient of variation and one as a standard deviation, and they are not the same figure. Labs report different ones and do not always say which, so a value from one report is not necessarily comparable with a value from another.

This one is worth being honest about. A large body of population research links a raised red cell distribution width to worse outcomes across a long list of unrelated conditions, and that literature is sometimes quoted as though it made the number something to optimise. It is an association in population data and not a target, and there is no evidence that acting on it in a healthy person changes anything. Read it as a flag to look at the iron and B12 markers, which is what it is genuinely good for.

In the post

Rises by about 8.4% at 24 hours at room temperature and 14.0% by 48, and stays flat refrigerated. So the one number here that already carries a warning against over-reading also carries a systematic false rise in a sample that has spent a day in transit.

White blood cell count

WBC

The total number of white cells, the immune system's circulating population. On its own it says little, and the differential below is where the information is.

It rises with infection, inflammation, physical stress, smoking and steroid medication, and falls in some viral illnesses and with certain drugs. A very high or very low count with no obvious explanation always warrants review, because the bone marrow conditions behind it are rare but matter.

The total is roughly the sum of the 5 types below, which is why a perfectly normal total can sit on top of a distribution that is not normal at all. Only roughly, though. On a modern analyser the total is its own measurement on a separate channel, and the differential is measured as proportions which are then multiplied by that total to give the absolute counts. So the 5 absolute counts inherit any imprecision in the total.

Three things move it more than people expect

The time of day. Sampled every 3 hours across a full day under standardised conditions, every white cell line oscillates and all of them peak around midnight, rising through the day. The total swings by roughly 1.6 x10⁹ per litre between its peak and its trough. Relative to the width of its reference range, that places the total white cell count among the most strongly 24-hour-cyclical numbers on a blood test, alongside iron and haemoglobin. It rests on one small study in young men, and older work places the neutrophil peak in the evening instead of at midnight, so treat the clock time as less settled than the size of the swing.

Hard exercise, by a lot. Straight after a marathon, 98% of runners have a raised white cell count, averaging 18.97 x10⁹ per litre with a neutrophil count of 15.69. That is roughly triple a typical resting value, in healthy people, with nothing wrong with them.

Smoking, and it is reversible. Among healthy adults, current smoking is the factor most strongly associated with a white cell count at or above 9.0 x10⁹ per litre, carrying more than 6 times the odds, ahead of a raised CRP, body mass index, male sex and younger age. Unusually, it raises all 5 lines, where a raised CRP raises only neutrophils and monocytes. Having previously smoked carries no association at all. Followed through annual checks, the count in smokers who quit fell from 5.642 to 5.374 x10⁹ per litre within a year and stayed down for at least 2 more, while it did not move in those who carried on.

Where the numbers come from

No professional body publishes a general healthy threshold for an adult. The population reference range is assay-dependent.

In the post

The total is reasonably robust and the differential is not, which is the worst possible combination. At 24 hours at room temperature the total was still where it started, while the neutrophil count had halved and the lymphocyte count had risen by three quarters. So a delayed sample can report a normal total sitting on top of a differential that is substantially wrong. This is the reason our lab names the white cell differential among the parameters a delay affects.

Neutrophils

The largest white cell population and the first responders to bacterial infection and tissue damage.

They rise quickly with bacterial infection, injury, hard exercise, physical stress and steroids. They are also the cell line most affected by chemotherapy and some immune-modulating drugs.

The exercise effect is larger than the phrase suggests. Straight after a marathon the mean neutrophil count is 15.69 x10⁹ per litre, roughly triple a typical resting value.

Where the numbers come from

The commonly used figure is an absolute neutrophil count below 1.5 x10⁹ per litre, with below 0.5 regarded as severe and 0.5 to 1.0 as moderate.

Those figures are published in named documents, which needs correcting, because they are often described as mere convention. The American cancer toxicity grading system sets its grades at those boundaries, a 2023 European guideline uses the same bands, and a 2026 expert guidance paper in a UK haematology journal states them directly.

What the named bodies do not do is agree. That same 2026 paper notes that for white adults a threshold of 1.8 x10⁹ per litre is adopted for the definition of neutropenia according to the World Health Organization. So there are at least 2 published figures for the same boundary, 1.5 and 1.8, and the usual caveat applies, which is that these are consensus landing points and not levels derived from outcomes.

And the thing this marker shows about reference ranges generally

Many people have a naturally lower neutrophil count with no increased risk of infection at all.

The cause is a single-letter variant in the promoter region of the ACKR1 gene, which disrupts a binding site for a transcription factor used only in developing red cells. The effect is to switch off expression of the Duffy antigen on red cells while leaving the same gene working normally everywhere else in the body. People with the resulting Duffy-null blood type commonly sit below the standard reference range while having entirely normal immune function and normal bone marrow.

This used to be called benign ethnic neutropenia. The 2023 paper behind the earlier figures below argued for replacing that term with Duffy-null associated neutrophil count, on the grounds that the old name attached a disease label to a healthy variant and tied it to ethnicity instead of to the genetic cause. The 2026 papers use a third name, ADAN but is also referred to as ACKR1 and DARC associated neutropenia.

The numbers, and they were updated in January 2026

The earlier figures put the median neutrophil count at 2,820 cells per microlitre in Duffy-null individuals against 5,005 in Duffy non-null, with about a quarter of Duffy-null individuals sitting below 2,000 and none of the non-null. Independent work in 2025 landed on almost exactly the same numbers, 2,801 against 4,959. Working from genotype instead gave medians of 3,270 against 4,830, with 17.3% of Duffy-null adults below 1,500 and 5.5% below 1,000.

In January 2026 much larger work replaced all of that, across 4 continents, and published actual reference intervals for Duffy-null individuals for the first time. The UK interval is 1,185 to 5,462 cells per microlitre for the neutrophil count and 3.1 to 8.8 x10⁹ per litre for the total white cell count. The 4 populations did not differ from each other.

The figure from that study that matters most is this. Measured against the reference ranges their own local labs were using, 26.0% of Duffy-null people in the UK cohort would have been classified as neutropenic. In the Saudi cohort it was 50.9%.

Why ethnicity is the wrong way to think about this

The variant is most common in people whose ancestors lived where malaria is endemic, because being Duffy-null is partially protective against Plasmodium vivax, which uses the Duffy antigen to enter red cells. Partially is the right word. Infections in Duffy-negative people are documented, a second species uses the same route, and a 2026 paper showed that species adapting to invade Duffy-negative cells anyway.

But the variant is not confined to any group, and the American Society of Hematology is explicit about the reasoning. Its own study documentation states that race and ethnicity are not biological facts but sociopolitical constructs, that genetic traits do not obey racial boundaries or geographic constructs, and that there should be extreme caution about using them as a proxy.

There is now hard data behind that. A 2025 analysis screening more than 1,400 clinical conditions across 3 large US datasets found the Duffy-null genotype reproducibly associated with changes in white cell count and with no disease outcome at all, and reported its prevalence as 68% in populations genetically similar to African reference populations, 14% in Middle Eastern, 12% in people identifying with more than one race, 7% Pacific Islander and 5% Hispanic. Its conclusion was that race is not an accurate predictor of Duffy-null status. The 2026 multinational study found no significant difference in neutrophil count between Black and non-Black Duffy-null participants, which is about as clean a demonstration as possible that the variant and not the ethnicity is doing the work.

The practical consequence is real, and it has been counted

People with this variant are routinely mislabelled as neutropenic, which leads to unnecessary investigation and, in some settings, to treatment being delayed or withheld.

Across late-stage trials in the 5 most common cancers, in the UK and US, 76.5% excluded people whose neutrophil count fell inside the Duffy-null reference range. For chemotherapy trials specifically it was 81.7%. Of the preferred curative-intent chemotherapy regimens, 53.5% contained dose reductions triggered by a neutrophil count inside that range, rising to 74.2% for lung cancer regimens. Those practices structurally exclude and under-dose people of African and Middle Eastern background.

A second worked example, from a different corner of medicine. Clozapine requires neutrophil monitoring because of a risk of drug-induced neutropenia. A 2025 study of schizophrenic on anti-psychotics with the Duffy-null variant showed neutrophils dropped below 1,500 cells per microlitre at least once over 7 weeks of weekly testing without taking clozapine at all compared to zero incidences in those without the variant.

Where this has reached UK practice is narrower than it sounds. A 2026 UK guideline on clozapine monitoring states that it is usual UK practice to lower the white cell and neutrophil thresholds for people confirmed to have the variant, and it recommends testing for the variant before clozapine is started. Outside of this, that interval is not what a general blood count report is measured against.

Units

The UK reports x10⁹ per litre and the US cells per microlitre. Multiply by a thousand, so 1.5 x10⁹ per litre is 1500 cells per microlitre.

In the post

Falls with delay, and steeply. At room temperature it is down to about half of baseline by 24 hours and a quarter by 48. When refrigerated it was still sitting at its starting value at 48 hours, so cold protects this line rather than just delaying the fall, though what cold does is not uniform across analysers.

So a posted sample systematically manufactures a low neutrophil count, for a marker where a genuinely low count is often nothing at all. A result below the range on a delayed sample could be the variant described above, or it could be the transit time, and the report cannot tell you which.

Lymphocytes

The white cells responsible for targeted immunity: T cells, B cells and natural killer cells. They handle viral infection and immune memory. How the immune response is organised is in Inflammation, in the Body Guide.

They typically rise in viral illness and fall with steroids, severe physical stress and some chronic conditions.

Exercise moves this line more than any other, and in both directions

During exercise the count rises, with natural killer cells increasing up to tenfold and CD8 T cells around two and a half fold. Then it falls, by 30 to 60%, reaching its lowest point 1 to 2 hours after finishing and returning to pre-exercise levels within 24 hours.

That is a large excursion in a healthy person, it is entirely physiological, and it means a sample taken in the hour or 2 after a hard session can show a genuinely low lymphocyte count for reasons that have nothing to do with immunity being suppressed.

It also has the largest proportional daily swing of the 5 white cell types, at roughly 1 x10⁹ per litre between its overnight peak and its daytime trough.

Where the numbers come from

No professional body publishes a healthy threshold for an adult. The population reference range is assay-dependent.

Both the absolute count and the percentage are reported, and the percentage is generally the less useful of the 2, because it moves whenever any other cell type moves. A falling lymphocyte percentage with an unchanged absolute count means some other line has gone up, usually the neutrophils. Every published action threshold in this area, for neutropenia, for lymphopenia, for a raised eosinophil count, is written as an absolute count, which is the strongest argument for reading the absolute figure.

Two qualifications on that. There is one place where a percentage is the defined measure, which is that a basophil count at or above 20% of white cells is a criterion for a phase of chronic myeloid leukaemia under one of the 2 current classifications. And on a modern analyser the absolute count is itself derived, by multiplying a measured proportion by the total white cell count, so it is not an independent measurement either.

In the post

Rises, and by a lot, reaching about 1.7 times baseline by 24 hours at room temperature. So neutrophils fall and lymphocytes rise in the same delayed sample, which is a pattern that looks meaningful and is not.

Monocytes

White cells that mature into macrophages once they enter tissue, clearing debris and dead cells and helping coordinate longer-running immune responses.

They are conventionally described as rising in chronic inflammation and in the recovery phase after an acute infection. That is the textbook account and it is reasonable. No modern primary source with figures behind it could be found, so treat it as the standard description and not an established finding.

Prolonged exercise does raise them. After a 308 kilometre continuous ultramarathon, monocytes rose alongside neutrophils.

Where the numbers come from

No professional body publishes a healthy threshold for an adult. The population reference range is assay-dependent. The percentage carries the same caution as lymphocytes.

Their daily swing is the second smallest of the differential in absolute terms with only eosinophils moving less.

In the post

The most stable of the 5 white cell types at room temperature, holding to about 24 hours and then falling to 88.3% of baseline by 48. One study using a different analyser found the monocyte percentage rising significantly by 48 hours even refrigerated, so the protection cold offers is not uniform across instruments.

Eosinophils

White cells involved in allergic responses and in defence against parasites.

They are commonly raised in hay fever, asthma, eczema and drug reactions, and they follow a daily rhythm, running higher overnight and lower in the morning. They are not special in it, though, because all 5 white cell types oscillate significantly across the day and all peak around midnight. Eosinophils have the smallest absolute swing of the 5.

Prolonged exercise lowers them. After a 308 kilometre ultramarathon the count fell alongside the lymphocytes, which is the opposite direction to the neutrophils in the same samples.

Where the numbers come from

There are published thresholds for this marker, which sets it apart from most of the differential. They come from consensus proposals and disease guidelines instead of from a general healthy range.

A 2012 international consensus proposal on eosinophilic disorders sets 0.5 x10⁹ per litre as the boundary of a raised count, 1.5 as the level defining true hypereosinophilia, and 5.0 as severe. In airways disease, asthma guidance uses figures of 150 and 300 cells per microlitre to guide treatment decisions. Those are disease-specific thresholds and not statements about a healthy person, and the consensus authors noted themselves that further work would be required to validate the level they proposed.

The population reference range is assay-dependent.

A single measurement may not be enough, and there is a figure for that

In anyone with an allergic condition the count varies with the season, by around 20% between July and January. Nasal polyps raised the count by 38%. Current smoking lowered it by 23%, which runs against the finding in the white cell count entry that smoking raises all 5 lines in a general health-check population, so both hold in their own settings.

The reproducibility figures from the same work matter more. Measured monthly for 10 months, of the people sitting in the 150 to 299 cells per microlitre band, only 44% stayed in that band for 7 of the 10 measurements. For those below 150 it was 59%, and for those at 300 or above 66%. The middle band is the least reproducible one, which is inconvenient, because it is the band the treatment thresholds sit in. A single measurement is not enough to place someone.

So 2 results taken months apart are not always comparable even when everything else about the person is the same.

In the post

Falls, and it is the first of the white cell types to go, measurably down by 8 hours at room temperature and roughly halved by 24. What refrigeration does to it is not settled.

Basophils

The smallest white cell population, involved in allergic and inflammatory responses through the release of histamine.

They are present in such low numbers that small absolute changes look dramatic as percentages.

Where the numbers come from

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

There is one published threshold, and it is a percentage instead of a count. The International Consensus Classification puts basophils at or above 20% of white cells among its criteria for accelerated-phase chronic myeloid leukaemia. The WHO classification has no equivalent, because its current edition recognises only a chronic and a blast phase and dropped the accelerated phase altogether. So the threshold is real but it belongs to one of the 2 classifications in use. That is the exception to the general rule that absolute counts carry the information and percentages do not.

Zero is a normal result

Basophils are present in numbers so small that the count sits at the edge of what the analyser can resolve, and a result of zero is a normal finding and not a deficiency. There is no recognised inherited or primary basophil deficiency. A low count can also follow steroids, acute infection or physical stress, but at this end of the scale the number is largely analytical noise in either direction.

How coarse this number actually is

This needs spelling out, because it is the least precise line on the whole blood count and that is not obvious from the report.

A 5-part differential typically classifies something on the order of 10,000 white cell events. At a true basophil fraction of around 0.5%, that means roughly 50 events are defining the result. Everything follows from that.

Against a manual count by a person looking down a microscope, automated basophil counts agree poorly while the other 4 types agree well. Where the other 4 lines track the manual count closely, basophils track it barely at all, and that holds across several analyser platforms. The automated basophil number is the least trustworthy figure in the differential.

And it is not the analyser's fault. Measured by a flow-cytometric reference method, the variation between labs was 0.3 to 0.7% for neutrophils, 0.5 to 0.9% for lymphocytes, 1.7 to 2.6% for monocytes, 3.0 to 7.9% for eosinophils and 3.8 to 10.4% for basophils. Basophils are the hardest of the 5 to count by any method.

In the post

Rises more than any other line, reaching about 2.6 times baseline by 24 hours at room temperature, and on a figure that imprecise by then it is unreliably wrong as well as wrong.

Platelet count

The cells responsible for clotting. They plug damaged vessel walls and start the clotting cascade, and what a damaged vessel wall looks like is in Heart and blood vessels, in the Body Guide.

What actually raises it

Inflammation, infection and tissue damage. In a consecutive series of people with a count of 500 x10⁹ per litre or above, the causes were tissue damage in 42%, infection in 24%, cancer in 13% and chronic inflammation in 10%.

Iron deficiency, with a caveat that matters here. It holds for established iron-deficiency anaemia, where around 28% of affected women have a raised platelet count. It does not appear to hold for low iron stores without anaemia, which is a much larger group. Across paired ferritin and platelet measurements in healthy blood donors there was no relationship between the 2 at all, and none appeared at any level of ferritin. That was a deliberate test of the textbook claim and it did not find it.

Exercise, briefly. Strenuous exercise produces a modest rise immediately afterwards which settles after a few hours. It is a reason not to give a sample straight after a hard session and not a lasting feature. Separately, platelet clumping in the circulation has been reported in healthy marathon runners 30 minutes after a race, which is a measurement hazard and not a physiological one.

Where the numbers come from

No professional body publishes a healthy threshold for an adult. The population reference range is assay-dependent.

A low count, and where bleeding risk actually begins

A low count raises bleeding and bruising risk, but not evenly across the range.

In newly diagnosed immune thrombocytopenia, the count that best separates those who bleed from those who do not is 20 x10⁹ per litre. Below 10, bleeding is dramatically more likely than above 20, by something close to 50 times. Between 10 and 19 it is around 5 times more likely. Anti-inflammatory painkillers and anticoagulants raise the risk independently of the count, anticoagulants by more than 4-fold for severe bleeding.

The familiar figures of 50, 30, 20 and 10 need placing correctly. They come from the British Society for Haematology's guideline on platelet transfusion, and they are thresholds for giving platelets before procedures or during bleeding, not the points at which bleeding starts. That guideline maintains 10 as the prophylactic threshold in reversible marrow failure, 50 for major surgery and severe bleeding, and 100 for neurosurgery, and it notes that the meta-analysis behind the figure of 10 may not have been powerful enough to detect an increase in bleeding risk of less than 50%.

A high count, which is a different story from the one usually told

In that same series, 87.7% of raised counts were reactive and 12.3% came from a bone marrow disorder. Essential thrombocythaemia accounted for 45% of the latter.

The finding that changes the framing is what happened to the 2 groups. In the reactive group, thromboembolic events were confined to the venous system and happened only where other risk factors were present. With no additional risk factor, a reactive raised platelet count carried no meaningful clotting risk. The clotting risk belonged to the marrow conditions.

Thrombocytosis is defined at 450 x10⁹ per litre, and essential thrombocythaemia requires a persistently raised count at or above that figure. The figure of 600 turns up in some local pathways but is not the figure the current definition rests on. For scale on the condition itself, incidence is about 1.5 per 100,000 per year, median age at diagnosis 59, around 90% carry an identifiable driver mutation, and arterial thrombosis occurs in about 11%, venous in 7% and bleeding in 8%.

So a single raised count rarely warrants much. A persistently raised one is what gets investigated.

The measurement problem, which is a real one for a posted sample

The anticoagulant used for a blood count, EDTA, works by stripping calcium out of the sample. That changes the shape of a receptor complex on the platelet surface and exposes parts of it that are normally hidden. In a minority of people, antibodies already present in their blood recognise the changed shape and stick the platelets together. Antibodies against that receptor are found in around 80% of cases, and platelets from people who lack the receptor entirely do not clump this way.

The analyser then counts each clump as a single platelet, or mistakes it for a small white cell, and reports a falsely low platelet count and sometimes a falsely raised white cell count. Nothing is happening in the person. People who have this have had it for years without consequence.

It is uncommon in the population, between about 0.03% and 0.3% of routine blood counts. It is common among low counts, which is the part that matters. In outpatients referred specifically for an unexplained low platelet count it accounted for 15 to 17% of cases, making it the second most common explanation. Where that has been looked at, the reported count averaged 42 x10⁹ per litre and the true count on a fresh sample averaged 208.

Whether a lab catches it depends on how the lab counts. The most common method, measuring the electrical impedance of each cell as it passes, is the most vulnerable, because a clump reads as one large platelet. The histogram it produces can show a characteristic irregularity, which is a clue an experienced operator may notice and not an automatic flag. Analysers that count platelets by fluorescence break the clumps up and get substantially closer to the truth. The reliable check is a person looking at a blood film, and the confirmation is a repeat in a different tube, either citrate or a magnesium-based one, in which the clumping does not occur.

Two things about timing make this worse for a posted sample. The clumping builds progressively over the first 4 hours after the blood is drawn, so the standard way of demonstrating it is to count immediately and again at 4 hours. A posted sample therefore arrives at close to maximum expression of the artefact. And where the rate came out as high as 17%, the reasons given were a lack of microscopic inspection of the blood film in primary care labs together with considerable delay in getting samples processed.

In the post

Genuinely contested, with published stability running anywhere from under 4 hours to 48, and the disagreement is itself the finding. The direction is usually downward, through clumping and adhesion. On top of that sits the clumping artefact above, maximally expressed by the time a posted sample arrives and with no blood film in a postal workflow to catch it.

MPV

Mean platelet volume

The average size of a platelet. Larger platelets are on the whole younger and more reactive.

It is often described as showing how fast the marrow is producing platelets. That job belongs to a different measurement.

The marker that actually tracks platelet production is the immature platelet fraction, which detects the residual genetic material inside newly released platelets. It is the platelet equivalent of a reticulocyte count.

Compared head to head against healthy people and against low platelet counts from 2 different causes, the immature fraction separates destruction from underproduction well, platelet distribution width does so less well, and average platelet size is the worst of the 3. In the people whose marrow was underproducing, average platelet size was no different from healthy controls at all. So average platelet size and the related indices should not be used on their own to work out why a platelet count is low.

Where the numbers come from

There is no published threshold for this marker, and unusually, there is no agreed way of measuring it either.

The call for a reference method was made in a paper published in 1984 and titled for the purpose, and it has never been answered. There is no reference method and no reference interval from any standards body. The value depends on the anticoagulant, the analyser and the time since the blood was drawn. In one study of the same donors the reference range came out as 7.2 to 10.8 femtolitres in EDTA and 6.1 to 9.5 in citrate.

Reviewed in 2026, almost every study reported an association with heart attack, stroke and thromboembolism, and the certainty of that evidence was then rated very low. The reason is that the cut-offs used ranged from 7.45 to 12 femtolitres, so the studies are not measuring the same thing. The conclusion was that the marker needs standardising before it is used. Lab scientists reviewing the same problem put it more bluntly, that the measurement difficulties make these indices little more than experimental.

So it is read alongside the platelet count, and it is best understood as a research marker that happens to come free with the blood count.

In the post

This is the first number on the blood count to fail, and it is not close. Platelets swell in EDTA, so the measured size starts climbing within the first hour, cold does not rescue it, and the direction is not even fixed, with the same blood rising 17% on one analyser and falling 22% on another.

So it has no reference method, no reference interval and no professional-body threshold, and it is the least dependable number on a posted blood count. Said plainly so that nobody reads anything into it.