HbA1c vs Fasting Glucose: What Each One Measures and Why They Disagree

Fasting glucose is one reading of where your blood sugar sat after an overnight fast. HbA1c estimates your average glucose exposure across roughly the previous 8 to 12 weeks. Because they measure different things, they disagree often, and the disagreement is largest in the band people care most about: prediabetes. In a population study in Vietnam, HbA1c classified 34.6% of adults as prediabetic against 12.1% by fasting glucose, and agreement between the two tests was only moderate (weighted kappa 0.55). If your two results point different ways, that is the normal behaviour of two different measurements, not a lab error.

Those are the American Diabetes Association 2026 criteria, and a diagnosis requires two abnormal results, either two tests from one sample or the same test repeated. One detail that surprises people: the WHO uses the same diabetes cut-offs but puts impaired fasting glucose at 110–125 mg/dL (6.1–6.9 mmol/L), so a fasting glucose of 104 mg/dL is prediabetes under ADA criteria and normal under WHO criteria. The number did not change. The committee did.

What does fasting glucose actually measure?

A fasting plasma glucose test measures the concentration of glucose in your blood after 8 or more hours without food, which in practice means first thing in the morning. Overnight, almost all of the glucose in your blood is glucose your liver produced, not glucose you ate. So a fasting result is mostly a statement about your liver: how much glucose it is releasing, and how well insulin is restraining it.

That makes it a narrow but sharp instrument. It is cheap, it is fast, it is standardised across labs, and it is the value that most treatment decisions have historically been built on. What it cannot tell you is anything about the other 16 hours of your day. Two people with an identical fasting glucose of 95 mg/dL can have completely different post-meal behaviour, and fasting glucose will never show it.

It is also a single moment, which means it carries all the noise of a single moment: an incomplete fast, a short night of sleep, a stressful commute, or the ordinary morning cortisol rise that nudges glucose upward. (For how much that rhythm moves, see cortisol across the day.)

What does HbA1c measure, and over what window?

Glucose in your blood attaches non-enzymatically to haemoglobin inside your red blood cells, and once attached it stays attached for the life of that cell. HbA1c reports the percentage of your haemoglobin carrying that sugar, so it is a cumulative record rather than a snapshot.

The usual shorthand is "the last 3 months", which is roughly right and slightly misleading. Red cells live about 120 days, but your circulating population is a mix of new and old cells, so the measurement is weighted toward recent weeks: the previous month contributes far more than the month before that. A change you made three weeks ago is already partly visible. A change you made 14 weeks ago is nearly gone.

The ADAG study gave HbA1c a unit people can reason about. Across 507 participants wearing continuous monitors and taking frequent fingersticks, HbA1c tracked average glucose closely (r = 0.92), and the relationship became a formula: estimated average glucose (mg/dL) = 28.7 × HbA1c − 46.7. In SI units, eAG (mmol/L) = 1.59 × HbA1c − 2.59.

Read that table carefully, because it is a common source of panic. An HbA1c of 6.0% converts to an estimated average glucose of ~126 mg/dL, which happens to be the number that defines diabetes when it appears as a fasting value. These are not the same quantity. eAG is a 24-hour average that includes every post-meal peak; a fasting value is the daily floor. An eAG of 126 mg/dL is entirely compatible with a fasting glucose in the 90s.

My HbA1c and fasting glucose disagree. Which one is right?

Probably both. Discordance is the rule, not the exception, and it has a direction you can read.

In the Vietnamese cohort study, of the 427 people whom fasting glucose placed in the prediabetes band, HbA1c classified 28.6% as having outright diabetes. Overall prevalence by HbA1c was 9.7% diabetes and 34.6% prediabetes; by fasting glucose it was 6.3% and 12.1%. The two tests were not ranking the same people.

The useful question is not which test wins but what the pattern means:

• Normal fasting glucose, elevated HbA1c. Your liver holds the line overnight but your average across the day is higher than that suggests. The usual explanation is post-meal excursions that a fasting draw never sees. It can also be one of the HbA1c artefacts below.
• Elevated fasting glucose, normal HbA1c. Either the fasting reading was noise (test conditions, one bad night, an incomplete fast) or your mornings are genuinely the worst part of your day while the rest is fine. Repeat the fasting test before concluding anything.
• Both elevated. The agreement is the signal. This is the combination worth acting on with a clinician.

What not to do is average them, or pick the friendlier one. They are different measurements in different units of different processes.

Which one predicts long-term risk better?

For people who do not already have diabetes, HbA1c has the stronger track record. In the ARIC cohort, 11,092 adults without diabetes or cardiovascular disease at baseline were followed for roughly 15 years. HbA1c and fasting glucose predicted future diabetes about equally well, but HbA1c was more strongly associated with cardiovascular events and death from any cause. Fasting glucose added little once HbA1c was accounted for.

Two honest caveats. First, this is an observational cohort: it shows HbA1c is the better marker of risk, not that lowering HbA1c by any particular means changes that risk. Second, "better on average across 11,000 people" is not the same as "better for you", which is the whole reason the artefacts in the next section matter.

If you want the earliest warning rather than the best risk marker, neither test is the first to move. Fasting insulin typically rises years before either glucose measure drifts, which is why it is worth pairing these with fasting insulin and the HOMA-IR calculation.

When HbA1c misleads

HbA1c is a measurement of red blood cells that we interpret as a measurement of glucose. Anything that changes red cells changes the number without your glucose changing at all.

• Red cell lifespan. Cohen and colleagues measured mean red cell age in people who were haematologically normal and found it ranged from 38 to 60 days in controls and 39 to 56 days in people with diabetes. Longer-lived cells accumulate more sugar, so they report a higher HbA1c at the same average glucose. The authors concluded the spread was large enough to matter clinically.
• Anaemia and iron status. Iron deficiency tends to raise HbA1c; haemolysis, recent blood loss and recovery from bleeding tend to lower it. So does anything that shortens red cell survival.
• Haemoglobin variants. Sickle cell trait, HbC, HbE and thalassaemias can interfere with some assays. If you carry a variant, the assay method matters and your lab should know.
• Pregnancy, advanced kidney disease, recent transfusion, erythropoietin. All shift the number independently of glucose.
• Population differences at the same glucose. In a study using continuous monitoring to fix mean glucose, HbA1c ran roughly 0.4 percentage points higher on average in Black than in White participants at the same measured mean glucose. Averages, not individuals, but enough that a single cut-off is not equally accurate for everyone.
• The average hides the shape. Beck and colleagues compared HbA1c against continuous monitoring in 387 people and found HbA1c both over- and underestimated mean glucose, sometimes substantially. Two people at 7.0% can be living quite different days.

None of this makes HbA1c a bad test. It makes it a test with a mechanism you should know, so that a result you cannot explain sends you to the mechanism rather than to a conclusion.

When fasting glucose misleads

Fasting glucose has no red cell problem. Its weakness is simpler: it is one number from one morning, and mornings vary.

• An imperfect fast. Milk in coffee, a late snack, or a fast shorter than 8 hours invalidates the comparison with your previous results.
• Short or broken sleep. A single poor night measurably reduces insulin sensitivity, and the fasting draw the next morning reflects it.
• Acute stress and illness. Both raise glucose transiently. A blood test during a cold is not your baseline.
• Time of day. The morning cortisol rise pushes hepatic glucose output up, so an 8 a.m. draw and an 11 a.m. draw are not interchangeable.
• Exercise before the draw. A workout changes glucose uptake acutely. Train after the blood test.
• Ordinary biological variation. Research on biological variation in HbA1c and glucose shows that repeating the same test in the same person reclassifies a meaningful share of people sitting near the prediabetes thresholds. A result of 101 mg/dL is not reliably different from a result of 99 mg/dL.

This is why the ADA requires two abnormal results before a diagnosis, and why one number near a threshold is a reason to retest rather than to change your life.

How should you track both?

A workable routine for someone who is not diabetic and wants to watch this honestly:

• Test both together, from the same draw. The pair is more informative than either one, and the direction of any disagreement is itself data.
• Keep the conditions fixed. Same lab, 10 to 12 hours fasted with water only, morning draw at a consistent time, no workout beforehand, a normal night of sleep. Changing the conditions changes the number without your body changing.
• Retest before reacting. Especially within 5 mg/dL of 100 or 126, or within 0.2 percentage points of 5.7% or 6.5%.
• Add fasting insulin once. If both glucose measures are normal, insulin is where an early problem would show first.
• Do not retest HbA1c sooner than 8 to 12 weeks after a change. Before that you are measuring the old you.
• Watch the trend, not the point. Three results over 18 months tell you something a single result never can.

This is the part that usually fails in practice, and not for lack of discipline: the results arrive as PDFs in an inbox, months apart, in units that changed when you changed labs. Logging each panel in BioTrakk (send a photo of the lab report, or type the values) turns them into one trend line per marker, with the test conditions attached, so you can tell a real move from a noisy morning. If you want to test an intervention against your own baseline rather than against a population average, that is what an n-of-1 experiment is for, and BioTrakk will return an honest inconclusive when the data does not support a verdict. Storing lab results over time is described in more detail on the blood test tracker page.

BioTrakk is not a medical device and this is not medical advice. HbA1c and fasting glucose are diagnostic tests: interpreting yours, and deciding anything on the basis of them, is a conversation with a clinician. Nothing here diagnoses, treats or prevents diabetes.