Blood sugar and metabolism

What insulin actually does at the cell

Glucose cannot cross a cell membrane on its own. It needs a transporter, which is a protein that sits in the membrane and lets it through. In muscle and fat those transporters are not permanently on the surface. They are held inside the cell, packaged and parked, waiting.

Insulin binds a receptor on the outside of the cell. That triggers a chain of signals inside, and the end of the chain is an instruction to move those parked transporters out to the surface. Glucose then flows in through doors that were already built.

So insulin does not push glucose into cells. It opens doors. And insulin resistance is not a shortage of doors or a shortage of insulin. It is a failure somewhere along the signalling chain between the receptor and the doors, so the same amount of insulin moves fewer of them.

Where the fat sits, and what the scans actually show

That signalling failure does not arrive at the same body weight in everyone. Cardiometabolic risk turns up at a lower body mass index in South Asian, Chinese, other Asian, Middle Eastern, Black African and African-Caribbean people, and UK guidance acts on it. NICE classifies overweight from a body mass index of 23 and obesity from 27.5 in all of those groups, against 25 and 30 for everybody else. Notice the shape of that. It is a single pair of numbers covering 6 very different populations rather than a figure worked out for each, and NICE presents it as a practical measure rather than a precise figure.

What the scans show is why. At the same body mass index, South Asian people carry more fat inside the liver, more fat inside skeletal muscle and less lean mass than white Europeans, and their fat cells are larger and fuller. The visceral excess is at its clearest in the leanest people, which is what people are pointing at when they say somebody looks thin and is metabolically loaded, and it narrows at the heavy end. Adjust for fat cell size and most of the difference in fasting insulin and HDL goes with it, which points at how much fat the safe storage compartment can hold rather than at how much fat there is in total.

Ordinary obesity sits on top of that rather than instead of it. The 2 are not alternatives, and a South Asian body carrying excess weight is carrying the storage problem as well.

It also does not travel across groups. In UK data African-Caribbean people had less estimated visceral fat than Europeans, not more, while both groups carried thicker trunk skinfolds. Writing that South Asian and African-Caribbean people alike carry more visceral fat is half wrong.

Body composition does not close the gap either. Adjusting for insulin resistance and trunk fat removed the extra diabetes risk in Indian Asian and African-Caribbean women in UK follow-up and left it untouched in the men, and there is no clear evidence that genetic factors contribute much of the excess. A large part of the difference in men is unexplained.

There is no ethnicity-specific waist number in UK guidance

The 90 cm waist quoted for South Asian men gets attributed to NICE constantly. NICE has never published it, and it dropped waist cut-offs in centimetres from its obesity guideline altogether, replacing them with a waist-to-height ratio of 0.5 that covers both sexes and all ethnicities. Keep your waist under half your height is the whole instruction, and it is the same instruction for everybody. Where the 90 cm figure does come from, and what it leaves out, is in Blood sugar, in the Biomarker Guide.

None of this is on a blood panel, which is why it sits here rather than with the markers. Every figure above came from imaging, from a fat biopsy or from a tape measure. No blood marker measures visceral fat, liver fat, fat in muscle or lean mass, and NICE says not to use bioimpedance in place of body mass index. The tape measure is the measurement.

Why glucose is the last thing to move

The body's response to that is to shout louder. The pancreas produces more insulin, more doors get moved, and glucose stays where it should be.

The relationship between the two is not a straight line. As sensitivity falls, the amount of extra insulin needed to hold glucose steady rises disproportionately. So for a long stretch, insulin output climbs steeply while the glucose number does not move at all. It only moves when output finally stops keeping up.

A normal fasting glucose does not mean the system is untroubled. It means compensation is still winning. By the time glucose is consistently high, that compensation has been running for years.

Whose bodies that sequence was measured in

The order is measured rather than assumed. Insulin sensitivity falls steeply across the 5 years before diagnosis. Beta cell output surges about 4 years out, holds briefly, then collapses. Glucose after a meal climbs from about 3 years out, and fasting glucose moves last of all. The people measured were almost all white and mostly male, so a white British sequence is standing in for everybody.

When it starts is not in doubt. Type 2 diabetes is diagnosed on average about 12 years earlier in South Asian people and about 10 years earlier in African-Caribbean people than in White European people, and at a lower body weight. So the years of compensation described above are running through people's 30s and 40s rather than their 50s, in a body that does not look the way most people picture. The figures behind that, and what the prevalence gap actually is once age and deprivation are allowed for, are in Blood sugar, in the Biomarker Guide.

The sensitivity is visible in childhood. Among UK children aged 9 and 10, the same increase in body fat raises insulin resistance further in South Asian children than in white European children, with African-Caribbean children in between. Whatever drives it is not waiting for adult life to begin.

The measurement that would catch it, and why almost nobody runs it

If insulin climbs for years while glucose sits still, the obvious move is to measure the insulin. You can. It is a fasting blood test, and paired with a fasting glucose it gives you a score called HOMA-IR, which estimates how hard your pancreas is working to hold that glucose where it is.

It does the one thing this section says nothing else does. It picks the problem up during the years of compensation, before glucose or HbA1c have moved at all. Somebody can have a fasting glucose and an HbA1c that both read comfortably normal and a HOMA-IR that clearly is not.

No major diabetes guideline recommends it for screening, and the reason is not that the idea is wrong. Insulin assays are not standardised between labs, so the same sample can produce different numbers in different places, and there is no agreed cut-off that holds across populations. The model also assumes your pancreas is still working properly, so it stops meaning much once diabetes is established. At least one study found it tracked poorly against the gold-standard method in a non-European population.

We do not sell it. Most private labs do, and a GP can request it, though plenty will ask why. If you want it, ask for fasting insulin alongside a fasting glucose on the same sample. The calculation costs nothing and there are calculators online.

Read it as a trend in yourself rather than against somebody else's threshold. Given the assay problem, that is the only sensible way to use it anyway.

Pregnancy runs the same sequence in months

Pregnancy induces insulin resistance on purpose. Placental hormones push insulin sensitivity down through the second half, the pancreas compensates, and glucose holds steady until compensation stops keeping up. Gestational diabetes is the failure described above, compressed from years into months.

Specific ethnic groups used to be named in the UK risk factors for it and are not any more. NICE listed South Asian, Black Caribbean and Middle Eastern family origin at the booking appointment, then deleted the names in December 2020 because the list was not exhaustive and missed groups out. The wording now is an ethnicity with a high prevalence of diabetes, with no list attached. That is a guideline stepping back from named groups as a proxy in public, and saying why.

London maternity data supports the retreat. Diabetes in pregnancy ran at about 4.6% in South Asian women, 3.7% in East Asian women, 3% in Black women and 1.5% in White women, and East Asian family origin was never on the deleted list. Weight behaves differently across the groups as well, since a body mass index above 30 covers about 35% of South Asian women who have diabetes in pregnancy against about 56% of Black women.

NICE also leaves the pregnancy body mass index trigger at 30 while using 27.5 outside pregnancy for the same groups. That is 2 guidelines from the same body carrying different numbers for the same person. Dropping the pregnancy figure to 27.5 would not repair it, because South and East Asian women at 27.5 already carry more than twice the risk that White women carry at 30.

What the profile looks at

Glucose is a snapshot of one moment, and given what the section above describes, it is the weakest thing on the panel for anyone who feels well. It is measured fasted because that at least standardises which moment.

If you want to see what your glucose actually does all day, there is a way, and it is not a blood test. A continuous glucose monitor is a small sensor worn on the arm for a week or two that reads glucose every few minutes. They were built for diabetes and are now sold to people who do not have it.

What they give you is the shape rather than the number. Which meals spike you and which do not, how long it takes to come back down, what a bad night's sleep does to the next morning. None of that appears on a fasting glucose or an HbA1c, and all of it is specific to you in a way no published guidance can be.

Two caveats. The readings sit slightly behind a finger prick, because the sensor reads the fluid between cells rather than blood itself. And glucose swings in people without diabetes are far more variable and far less meaningful than the marketing around these devices suggests. Treat it as a fortnight of education about your own responses rather than as a diagnostic.

HbA1c is more useful because it cannot be fixed by one careful morning. Glucose attaches to haemoglobin inside red blood cells and stays attached for the life of the cell, so the proportion that has been modified reflects the glucose those cells have been sitting in.

Why recent weeks count for more

The weighting is not a quirk of the assay. It falls straight out of the fact that your red cells are not all the same age.

At any moment you are carrying cells of every age from brand new to about 4 months old, with the oldest being continuously removed and replaced. Glycation builds up over a cell's life, so the oldest cells carry the most. But there are far more cells that have been alive through the last month than cells that have been alive through all 4. Most of the cells present today were already here a month ago, while only the few oldest survivors were here 4 months ago.

So recent glucose is represented in every cell in the sample, and older glucose only in the ones that have lasted. Roughly half the number comes from the preceding month. That is why an HbA1c responds to a change in the last few weeks but understates how big the change was.

The numbers, and which country's numbers they are

These thresholds are national guidance rather than anybody's house opinion, so here they are in full. In the UK, HbA1c is reported in millimoles per mole, and the bands come from the National Institute for Health and Care Excellence (NICE). The WHO sets the same diabetes line but has said there is not enough evidence for it to take a formal position on anything below that line. The bands are:

Below 42 is normal. 42 to 47 is non-diabetic hyperglycaemia, usually called prediabetes, meaning glucose is higher than it should be but below the diagnostic line. 48 or above meets the diagnostic threshold for type 2 diabetes, which in someone without symptoms requires a second test to confirm.

The band boundaries are not the same everywhere, and this catches people out. American guidance sets the lower boundary at 39 rather than 42. So a result of 40 is normal in Britain and prediabetes in the US. Same blood, same assay, different label. Bear that in mind before reading an American article about your own result.

When HbA1c lies, and it lies in both directions

HbA1c depends on red cells living a normal length of time. That dependency is behind most of the ways it misleads, and the two directions are not equally well known. Some haemoglobin variants throw the measurement off in a different way, by interfering with the assay itself.

Anything that shortens red cell life makes it read low. The cells are removed before they have accumulated their share, so the average comes out under the true glucose. Haemolysis, recent significant blood loss and a recent transfusion all do this.

Anything that makes red cells older on average makes it read high, and this is the half that usually goes unsaid. Iron deficiency is the common one. Cells survive longer, they accumulate more glycation, and the HbA1c rises without glucose having changed at all. Correcting the iron deficiency brings the HbA1c back down on its own. B12 and folate deficiency work the same way.

The practical consequence matters. Iron deficiency is common, particularly in menstruating women, and it can lift an HbA1c from the normal band into the prediabetes band in somebody whose glucose handling is entirely fine. Which is a good reason to read an HbA1c next to a full blood count and a ferritin rather than on its own. The Iron section explains why the ferritin would already have fallen long before the blood count showed anything.

Why the panel contains what it contains

The Metabolic Risk panel puts lipids, ApoB, hs-CRP and a full liver profile alongside these two, and that is not padding.

Insulin resistance rarely travels alone. It arrives with rising triglycerides and falling HDL, with fat accumulating in the liver, and with low-grade inflammation. Which is why a raised ALT or a high-normal GGT in the liver section, a triglyceride level in the lipid section and an hs-CRP in the inflammation section can all be describing the same underlying process from different angles. Reading them as 4 unrelated findings is the most common way this gets missed.

What changes when this is under strain

Early on, very little. Later, increased thirst, passing urine more often, tiredness, blurred vision, slower wound healing and recurrent infections.

The years of silence before that are the years worth measuring in.

How much movement, and whether the dose differs

The UK Chief Medical Officers looked at this and declined to vary the target. Their guidelines state that the data on physical activity and ethnicity does not suggest the relationship between activity and health varies by ethnicity, so there is no reason to vary the guidelines by it. The target is 150 minutes a week of moderate activity or 75 of vigorous activity, plus muscle strengthening on at least 2 days, for everybody.

Set against that, UK work puts the equivalent dose higher in South Asian adults, at roughly an extra 10 to 15 minutes a day to reach the cardiometabolic risk profile a white European reaches at 150 minutes. That is not UK policy, it has not been confirmed over time, and it comes from a small body of work. Pulling the other way, activity in UK children tracked just as strongly with insulin, body fat and inflammation in South Asian and African-Caribbean children as in white European children, so the same activity appears to buy the same benefit and the disagreement is only about how much of it is needed.

What the numbers on a panel do and do not say is a separate question from the body underneath them, and Blood sugar in the Biomarker Guide covers how ethnicity changes who is tested without changing where the line sits.