Glycation and HbA1c: What That Number Actually Measures

Your HbA1c is not a blood sugar reading. It is a slow stain. Glucose drifting past a red blood cell bumps into hemoglobin and simply sticks to it — no enzyme, no permission, no regulation, just chemistry and time. A red cell lives about 120 days and cannot wash the sugar off, so by the time it dies it is carrying a permanent record of every glucose it swam through. Draw a tube of blood, measure the average coating across the whole population, and you have read a three-month diary. Press play and watch the diary being written — then break it.

Try this: start on Diabetes, let the cells go brown, then hit Improved today. Glucose drops to 100 mg/dL instantly — and the A1c barely moves. Press ⏩ Skip 30 days three times and watch how long the old sugar-coated cells hold the number up. Then try Short red-cell life, where the number stops telling the truth altogether.

The Science in Plain Language

Glycation is chemistry, not biology — nobody is in charge of it

Almost everything your body does to a sugar molecule is supervised. An enzyme grabs it, holds it in a precise pocket, does one specific thing to it, and lets go. Glycation is the exception. It is what happens when nobody is supervising.

A small fraction of glucose in your blood is floating in its open-chain form, and that open chain ends in an aldehyde group — a reactive carbon. Bump that reactive carbon into a free amino group on a protein and the two simply join. For HbA1c specifically, the amino group is the N-terminal valine on the beta chain of hemoglobin. The first bond formed is a Schiff base, and it is loose: it can come apart again within hours. But if it survives, it quietly rearranges into a far more stable form called an Amadori product (for hemoglobin, a ketoamine — fructosyl-valine). That rearrangement is essentially a one-way door. The sugar is now welded on for the rest of that cell's life.

Three consequences follow, and they explain nearly everything odd about the A1c test. First, glycation happens to everyone, always — a person with perfect metabolic health still runs an A1c around 5%, because they still have glucose in their blood. Second, the rate is set by nothing more than concentration and time: twice the sugar, twice the sticking. Third, there is no off switch, no feedback loop, no repair enzyme standing by to pluck the sugar back off. Your body cannot decide to glycate less.

Why the red cell's 120 days turns sugar into a three-month average

A red blood cell is unusual: it has no nucleus, no mitochondria, and no ability to make new protein. It is a bag of hemoglobin, sealed at birth in the bone marrow, that circulates for about 120 days before the spleen and liver take it out of service. It cannot repair itself and it cannot swap out stained hemoglobin for fresh.

So every red cell is a little tape recorder. A cell born this morning has essentially zero glycated hemoglobin. A cell born four months ago has been marinating the whole time and carries the most. Draw a tube of blood and you capture the entire population at once — newborns, teenagers, and pensioners together. The lab measures what fraction of all the hemoglobin in that tube carries the Amadori adduct, and reports it as a percentage. That is HbA1c.

Now the part almost nobody is told: the three months are not weighted evenly. Ask which calendar days contributed to today's number and the answer depends on how many cells were alive to record them. Every single cell in the tube was alive last week, so last week is recorded by 100% of the population. But four months ago, only the oldest cells — the small slice about to be retired — existed. Days that far back are recorded by almost nobody. The result is a lopsided average that leans hard on the recent past. The classic teaching figure is that roughly 50% of your A1c comes from the previous 30 days, about 25% from days 31–60, and the remaining 25% from days 61–120. The animation computes this live from the cells on screen and lands in the same neighbourhood (its simplified uniform-age population gives a little over 40%). Either way, the headline is the same: your A1c is mostly about last month.

How to read the number: the ADA bands and the eAG conversion

The American Diabetes Association thresholds are simple and worth memorising:

• Below 5.7% — normal.
• 5.7% to 6.4% — prediabetes. Not a diagnosis of diabetes, and not nothing either.
• 6.5% or higher — diabetes, confirmed on a second test unless blood sugar is unequivocally high with classic symptoms.

The problem with a percentage is that it means nothing to a person who checks their blood sugar in mg/dL. That gap is what the ADAG study (A1c-Derived Average Glucose, published in 2008) closed. Researchers put people on continuous glucose monitoring plus frequent fingersticks for three months, measured their true average glucose, and fitted it against their A1c. The line is remarkably straight:

eAG (mg/dL) = 28.7 × A1c − 46.7
or, in mmol/L: eAG = 1.5944 × A1c − 2.5944

Run it and the number stops being abstract. An A1c of 6.0% is an average glucose of about 126 mg/dL. 6.5% is about 140. 7.0% is about 154. 8.0% is about 183. 9.0% is about 212. Going the other way, if your meter has been averaging 180 mg/dL you should expect an A1c near 7.9%. The readout panel on the animation shows both directions at once, which is the point: the A1c gauge and the glucose gauge are the same information in different clothes — except when they are not, which is the next two sections.

One more unit you may see on a lab report, especially outside the United States: mmol/mol (the IFCC standard). The conversion is mmol/mol = (A1c% − 2.15) × 10.929, so 6.5% is 48 mmol/mol and 7.0% is 53 mmol/mol.

The lag: why a good week does not show up, and why a good month barely does

This is the single most useful thing on this page, and the reason the Improved today button exists. Suppose you change everything overnight — the food, the walking, the medication — and your glucose drops from an average of 200 mg/dL to 100 mg/dL starting today. What does your A1c do?

Almost nothing. Not because the change did not work, but because the evidence is physically still in your bloodstream. Every red cell you own was made before today and is already carrying its stain. Those cells have to age out and be replaced one by one, over 120 days, before the population average can reflect the new reality. Press ⏩ Skip 30 days on the animation and you will see a real fall — the last-30-days bar is the heaviest one, so the first month does most of the work — but you will also see that the number does not arrive at its new home for months.

Read that in both directions. It means a bad fortnight will not ruin a good A1c, and it also means an excellent fortnight cannot rescue a bad one. It means the A1c you are looking at today is grading work you did in the spring. And it means that when a treatment changes, re-testing at six weeks tells you which way things are heading, but only the three-month test tells you where you actually landed.

What A1c cannot see: a flat 7% and a violent 7% look identical

An average destroys information, and A1c is an average. Consider two people who both come back at exactly 7.0%. The first sits between 130 and 175 mg/dL all day, every day. The second spends part of each morning at 50 mg/dL, shaky and sweating, and part of each evening at 280. Their averages are the same. Their A1c is the same. Their lives are not remotely the same, and neither is their risk — hypoglycemia has its own dangers, and there is good reason to think that large swings do damage that a steady mean does not capture.

This is exactly the gap that continuous glucose monitoring fills. A CGM reports time in range — the percentage of the day spent between 70 and 180 mg/dL — along with time below range and glucose variability. The widely used target for most non-pregnant adults with diabetes is more than 70% of the day in range with less than 4% below 70 mg/dL. CGMs also produce a Glucose Management Indicator (GMI), calculated as GMI(%) = 3.31 + 0.02392 × mean glucose in mg/dL. GMI is an A1c-like number derived purely from sensor data, and it frequently differs from the lab A1c by a few tenths — which is not an error in either test. They are measuring different things: one is the fraction of glycated hemoglobin, the other is the arithmetic mean of interstitial glucose.

None of this makes A1c obsolete. It is cheap, standardised worldwide, needs no fasting, and predicts long-term complications extremely well at the population level. It is simply not a substitute for knowing the shape of your day.

When A1c lies — and which direction it lies in

Because A1c is a product of glucose and red-cell lifespan, anything that changes how long red cells live corrupts the result. The animation makes this literal: switch to Short red-cell life and glucose stays at 200 mg/dL while the reported A1c collapses into the normal range, because the cells are being replaced before they have time to accumulate a stain. Toggle Iron-deficiency anemia and the opposite happens.

• Hemolysis, or any shortened red-cell survival — falsely LOW. Younger average cell, less time to glycate.
• Recent significant blood loss, or the recovery burst after treating iron or B12 deficiency — falsely LOW, from the flood of brand-new cells.
• Recent blood transfusion — unreliable. You are now partly measuring the donor's blood sugar; the result is pulled toward the donor's value.
• Untreated iron-deficiency anemia — falsely HIGH, typically by a few tenths of a percentage point. The leading explanation is that the average circulating red cell is older than normal, though changes in glycation itself may contribute. Treating the iron deficiency lowers the A1c without anything happening to blood sugar.
• B12 or folate deficiency anemia, and asplenia (no spleen) — falsely HIGH, again through longer average red-cell age.
• Pregnancy — runs LOW, from faster red-cell turnover and expanded plasma volume. A1c is not the recommended test for diagnosing gestational diabetes.
• Chronic kidney disease, especially on dialysis or erythropoietin — usually falsely LOW: red-cell survival is shortened and EPO floods the circulation with young cells. Some older assays could read high in uremia because of carbamylated hemoglobin.
• Hemoglobin variants — sickle trait, HbC, HbE, HbD and others. Most modern assays handle sickle trait correctly, but not all methods do, and the NGSP publishes which assays are affected by which variant. In sickle cell disease, red-cell survival is genuinely shortened, so the A1c is low for real physiological reasons rather than assay interference.

The practical rule: if the A1c and the meter or CGM disagree persistently and substantially, believe the disagreement and look for one of the reasons above. A fructosamine or glycated albumin test, which reflects the previous two to three weeks and does not depend on red cells at all, is a common way to break the tie.

AGEs and the long game: the proteins that never get replaced

Hemoglobin gets a fresh start every four months. Most of your structural protein does not. Skin collagen turns over on a scale of years; the collagen in cartilage lasts longer still; and the crystallin proteins in the lens of your eye are, for the most part, the ones you were born with. On those proteins, glycation does not reset — it accumulates for decades.

And it does not stop at the Amadori stage. Given enough time, glycated proteins undergo further oxidation and rearrangement into a family of compounds collectively called advanced glycation end-products (AGEs): carboxymethyl-lysine, pentosidine, glucosepane and others. Several of them are cross-links — molecular staples that tie one collagen fibre to its neighbour. Glucosepane is the most abundant such cross-link found in human tissue.

A cross-linked collagen network behaves differently. It is stiffer and less elastic, which in an artery means a wall that cannot cushion the pulse and a rising systolic pressure; in the kidney it contributes to a thickened glomerular basement membrane and mesangial expansion; in the lens it contributes to clouding; in skin it shows up as reduced elasticity and can even be measured non-invasively as skin autofluorescence. The fourth panel of the animation shows this as amber ticks that appear between collagen fibres and never leave, straightening the fibres and narrowing the vessel.

AGEs also do something active. They are recognised by a cell-surface receptor called RAGE (receptor for advanced glycation end-products). Docking there switches on inflammatory signalling through NF-κB, producing oxidative stress and inflammatory cytokines. So the tissue is not just mechanically stiffer; it is being chemically nagged. That combination — stiffened matrix plus chronic low-grade inflammatory signalling, concentrated in the small vessels — is a large part of why long-running high blood sugar damages the retina, the kidney and the peripheral nerves in particular. Those tissues depend on delicate capillary beds that stiffen and leak.

The epidemiology matches the chemistry. In the UKPDS 35 observational analysis of people with type 2 diabetes, each 1 percentage point lower HbA1c was associated with roughly a 37% lower rate of microvascular complications, about a 21% lower rate of diabetes-related deaths, and about a 14% lower rate of myocardial infarction. In the DCCT trial in type 1 diabetes, intensive glucose control cut the onset of retinopathy by roughly three-quarters compared with conventional treatment. There is no threshold below which the relationship suddenly appears — it is a gradient, which is why moving from 9% to 8% is worth doing even if 7% is not yet in reach.

What actually moves the number — and how long to wait before re-testing

Only one thing changes your A1c: changing how much glucose is in your blood, and keeping it changed. Everything that lowers average glucose — the composition and timing of meals, movement after eating, weight change, sleep, treating an infection, and whatever medication your clinician prescribes — moves the number, and nothing else does.

The waiting period is not arbitrary. Because the red cell population needs to turn over before the average can settle, the standard interval is about three months. The ADA suggests testing roughly quarterly for people whose therapy has changed or who are not at goal, and about twice a year for those who are stable and at target. Testing at six weeks after a change is reasonable if you want an early direction of travel, but it will systematically understate the eventual improvement, so do not be discouraged by it. If you need feedback faster than that, fructosamine (two to three weeks) or a CGM (immediately) are the right tools — not a repeated A1c.

Two myths worth retiring

Myth 1: "I'll eat carefully for a few days before the blood draw." This does not work, and it cannot work. A1c is not a snapshot of today's blood sugar. There is no fasting requirement for the test precisely because what you ate this morning, yesterday, or over the past week has almost no measurable effect on the fraction of hemoglobin that was glycated over the past four months. Skipping breakfast changes your fasting glucose; it does not change your A1c by a detectable amount. The corollary is more encouraging than the myth: a stressful week, a holiday, or one bad weekend will not wreck the result either.

Myth 2: "My A1c is good, so my control is good." Not necessarily. A comfortable-looking 6.8% can be a steady 6.8%, or it can be the arithmetic midpoint between recurrent overnight hypoglycemia and daily post-meal spikes above 250 mg/dL. Those are different clinical situations with different risks, and the A1c alone cannot distinguish them. If you have hypoglycemia symptoms, wide swings, or any of the red-cell conditions listed above, a good A1c is a question, not an answer — pair it with time-in-range data or a fructosamine before concluding anything.

↑ Back to the animation

Connections

• All Interactive Visualizations
• Hemoglobin A1C — the lab test
• Fructosamine — the two-to-three-week alternative
• Continuous Glucose Monitoring — time in range
• A1c, Fructosamine and Glycation Markers
• Prediabetes
• Type 2 Diabetes
• Diabetes
• Diabetic Retinopathy
• Peripheral Neuropathy
• Carnosine — and the anti-glycation claims
• Animation: Blood Sugar & Insulin
• Animation: Insulin Signalling & GLUT4
• Animation: Erythropoiesis & EPO
• Animation: The Hemoglobin–Oxygen Curve
• Animation: SGLT2 & Kidney Glucose
• Animation: Oxidative Stress & Antioxidants