Fasting glucose, HbA1c and insulin measure three different things and they disagree more often than anyone tells you. What each one sees, what it misses, and which of them a trial has ever moved.
Kratos Natural · Domain: Blood work & testing
1. How is blood sugar measured?
They sit under each other on the report as though they were three phrasings of one question. They are not.
Fasting glucose is how much sugar is in your blood at one moment after a night without food. A photograph.
HbA1c is the proportion of your haemoglobin that has had sugar attached to it. That attachment is slow and permanent, and a red cell lives around three months, so the figure is a weighted average of that period. A long exposure rather than a photograph.
Insulin is not the sugar. It is the effort. It is what your pancreas is spending to hold that sugar where it is. Two people with identical glucose can be far apart here, and that gap is the entire point of the chapter.
Here is why the order matters. Your body defends blood glucose hard, and it will spend increasing amounts of insulin to do it. So the photograph keeps looking fine long after the cost of producing it has risen. Glucose only moves once the system stops keeping up, which is late. That is why a normal fasting glucose is less reassuring than it appears, and why insulin gets a chapter even though your report almost certainly does not carry it.
2. What is fasting glucose?
The fairest way to value a number is to ask what it predicts in a dataset too large to argue with.
The Emerging Risk Factors Collaboration pooled 102 prospective studies, 698,782 people and 52,765 vascular events. Having diabetes roughly doubled the risk of coronary heart disease (hazard ratio 2.00, 95% CI 1.83 to 2.19) and of ischaemic stroke, and adjusting further for lipids, inflammation or kidney markers barely changed it.
Then the part relevant to most readers here, the people without diabetes. In them the relationship between fasting glucose and vascular disease was non-linear and modest. Between 3.90 and 5.59 mmol/L there was no significant association at all. Above that it rose gently: 1.11 at 5.60 to 6.09, and 1.17 at 6.10 to 6.99.
And then the sentence that should change how you read your own report. In people without diabetes, adding fasting glucose to the conventional risk factors did not significantly improve prediction of vascular disease[1].
So fasting glucose is a good test for finding diabetes and a weak predictor of heart disease in somebody who does not have it. If you want one number about your arteries, it is in the lipids guide and it is called apoB.
Two practical notes, because this figure moves more than people expect. A poor night's sleep, an infection or a stressful morning all push it up with nothing having changed about your metabolism. And fasting means fasting. Coffee with milk counts.
3. What is HbA1c?
Sugar sticks to proteins. That is chemistry rather than pathology: the more glucose passes by, the more of it stays attached. Haemoglobin is merely the convenient protein to measure it on, because it lives inside a cell with a known shelf life.
That makes HbA1c immune to everything that makes fasting glucose jump. No fasting required, one bad night invisible, one holiday invisible.
What it is worth comes from the ARIC study: 11,092 adults with no diabetes and no cardiovascular disease, followed for years. Against a reference group at 5.0 to 5.5%, the hazard ratio for a later diabetes diagnosis was 1.86 at 5.5 to 6.0%, 4.48 at 6.0 to 6.5%, and 16.47 at 6.5% or above. For coronary heart disease the same bands ran to 1.78 and 1.95.
The result that settles the comparison is what happened under adjustment. Those associations survived adjustment for fasting glucose. The association between fasting glucose and cardiovascular disease did not survive adjustment for HbA1c[2].
Of the two numbers on your report, HbA1c carries more information. One caveat from the same paper is rarely quoted alongside it: against all-cause mortality the curve was J-shaped, rising slightly at the low end too. That is not a reason to distrust a low HbA1c. It is a reason not to say "lower is always better" in this chapter, which is a thing this site tries not to do anywhere.
The target shown here is under 39 mmol/mol, with 39 to 47 as the band the ADA calls prediabetes[3].
4. What if your glucose and HbA1c disagree?
This is the chapter the guide exists for, because it happens constantly and almost nobody is told about it.
At the same average glucose, some people glycate more haemoglobin than others. The difference has a name, the haemoglobin glycation index, and it is a stable property of the person rather than of the week.
Its size has been measured. 3,106 people without known diabetes had both an oral glucose tolerance test and an HbA1c. Judged by the tolerance test, the low-mismatch and high-mismatch groups had a similar spread of normal, prediabetes and diabetes. Judged by HbA1c alone, they did not. The low-mismatch group came out 97% normal and 0% diabetes. The high-mismatch group came out 13% normal, 77% prediabetes and 10% diabetes[4].
The same glucose handling. A completely different label.
The authors' conclusion is the instruction: HbA1c alone produces under-diagnosis in one group and over-diagnosis in the other, and fasting glucose or a tolerance test should accompany it.
Which gives you the single most useful thing in this guide. Have them drawn together. If they agree, you have your answer. If they disagree, you have something better than an answer: you know that your HbA1c reads differently from the table for you specifically, and that stays true for the rest of your life.
5. When is HbA1c wrong?
Separately from glycation differences, there is a set of situations where the number actively misleads, and the reason is always the same one. HbA1c is a measurement about red blood cells, so anything that changes red blood cells changes it.
Iron deficiency is the common one, and it pushes the wrong way. In a study of 62 people with iron deficiency anaemia against 60 healthy controls, HbA1c was clearly higher in the anaemic group (5.51% against 4.85%). After three months of iron supplementation it fell to 5.04%. Of twenty participants whose pre-treatment HbA1c sat in the prediabetes range, fourteen returned to the normal range[5].
A small study, and the direction matches the mechanism: cells that survive longer accumulate more sugar. The practical lesson is large enough to state plainly. If you are anaemic or your ferritin is low, get that corrected before an HbA1c is used to decide anything. Iron and ferritin are on the same screen here for this reason among others.
The opposite works too. Anything that shortens red cell survival lowers HbA1c falsely: haemolysis, a recent bleed, a transfusion, advanced kidney disease, and pregnancy.
Then the assay itself. A haemoglobin variant can interfere with the measurement, and how much depends on which analyser your lab runs. A comparison of six commonly used HbA1c methods found that the same variant produces a distorted result on some methods and not on others[6]. Variants such as HbS, HbC and HbE are most common in people with African, Mediterranean, Middle Eastern or South East Asian ancestry. If you know you carry one, say so to whoever orders the test.
6. What does insulin tell you?
The least assertive chapter here, on purpose.
Fasting insulin measures the effort behind a normal glucose. It moves earlier than glucose does, and it is mechanistically the thing you would most like to know.
It is still not on your report, and the reasons are respectable. Insulin assays are not standardised the way HbA1c is, so the same tube of blood produces different numbers on two machines, which makes a universal threshold impossible. No guideline sets a target. And no randomised trial has shown that anyone does better when a clinician steers by it.
Which is why this app carries no range for it. Not out of timidity: a range we invented sitting next to a real value is worse than no range, and the rule on this site is that a target names the guideline behind it. There is no guideline to name.
What it is good for is a line over time, from one lab. Two fasting insulins from the same analyser a year apart tell you something real. A single insulin value held against a table found online tells you nothing, and those tables tend to live on sites that sell the test.
7. What is HOMA-IR?
HOMA-IR is arithmetic on your fasting glucose and fasting insulin, described in 1985 as a model for estimating insulin resistance and beta cell function from exactly those two numbers[7].
What it can do: track you against yourself over time, provided both inputs come from the same laboratory.
What it cannot do: compare you to anyone else, or to a cut-off from the internet. It inherits chapter 6's problem completely, because putting a non-standardised number into a formula produces a non-standardised output. It is also a model built in an era of different assays, designed for describing populations rather than for judging whether your result is acceptable.
A line, not a verdict.
8. What do tolerance tests and sensors add?
Two hours in a waiting room after a sweet drink, a sample before and a sample after. Nobody enjoys it and it sees the most.
Chapter 1 explains why. A fasting value measures a system at rest. HbA1c measures its average. Neither measures how well the system copes when something actually arrives, and coping is the first function to go. Somebody with a respectable fasting glucose can be well above the threshold two hours after the drink, and that sequence is the usual sequence rather than the exception.
It is not a test to request casually. It appears when the other two contradict each other, in pregnancy, and when the symptoms and the numbers do not match. This chapter exists so that you know why your doctor might propose it when your report looks fine.
On continuous glucose monitors, since the question always arrives: a sensor shows you what a given meal does to you, and that is genuinely instructive. What does not exist is a threshold validated against outcomes in people without diabetes, or a guideline that names one. Treat what you see as education about your own breakfast rather than as a finding. This site prints no sensor thresholds for the same reason.
9. What counts as prediabetes?
The ADA criteria, which are also what this dashboard uses[3]:
| Normal | Prediabetes | Diabetes | |
|---|---|---|---|
| Fasting glucose | 3.9 to 5.5 mmol/L | 5.6 to 6.9 | 7.0 or above |
| HbA1c | under 39 mmol/mol | 39 to 47 | 48 or above |
Three things that appear on no report.
A diagnosis is never one measurement. One value across a line calls for a second sample, usually on another day. That is not bureaucracy. These numbers fluctuate, and a diagnosis follows somebody around.
Prediabetes is a threshold, not a disease. It marks a range where risk is raised, and most people in it never develop diabetes. It is a warning rather than a sentence, which is exactly why chapter 10 matters: this is the range where intervening works best.
And check your units. Dutch reports give glucose in mmol/L and HbA1c in mmol/mol. American sources use mg/dL and percentages. An HbA1c of 39 mmol/mol is 5.7%. Before taking fright at a number online, look at what is written after it.
10. Which blood tests should you ask for?
The lipids guide had to report that lifestyle moves those numbers less than people hope. Here it runs the other way, and this is the most cheerful chapter in the whole book.
The Diabetes Prevention Program. 3,234 people with raised fasting and raised post-load glucose, randomised to placebo, metformin, or a lifestyle programme aiming at at least 7% weight loss and at least 150 minutes of activity per week. Over an average 2.8 years, diabetes incidence was 58% lower in the lifestyle group (95% CI 48 to 66) and 31% lower with metformin. To prevent one case over three years, 6.9 people had to take part in the lifestyle programme[8].
Seven people. Elsewhere in this book the numbers needed to treat run to 56. This is among the largest and best-supported effects in lifestyle research, and it beat the drug it was tested against.
And once it is diabetes, DiRECT. 306 people diagnosed within the previous six years, in ordinary primary care practices. The intervention was substantial: antidiabetic and antihypertensive drugs withdrawn, a formula diet of roughly 850 kcal per day for three to five months, food reintroduced in steps, then structured support for maintaining the loss. At twelve months, 46% of the intervention group were in remission against 4% of controls.
The breakdown underneath is the real finding. Remission occurred in none of the 76 people who gained weight, 7% of those who lost 0 to 5 kg, 34% at 5 to 10 kg, 57% at 10 to 15 kg, and 86% of those who lost more than 15 kg[9].
So it is not the diet that works. It is the weight loss. And it belongs under supervision: stopping medication and eating 850 kcal a day are not things to begin on your own.
Training, measured separately. In people who already have diabetes, structured exercise lowered HbA1c by 0.67% against controls across 47 trials and 8,538 patients, rising to 0.89% above 150 minutes per week[10]. That is in the territory of a medication.
The thread through all of it: this panel responds. More than the lipid panel does, and far more than your Lp(a) ever will.
11. Which results need a doctor?
This week, not next month:
- Passing a lot of urine, constant thirst, unexplained weight loss, or blurred vision, and particularly those together. That combination belongs in front of a doctor now.
- A fasting glucose of 7.0 mmol/L or above, or an HbA1c of 48 or above. Not a number to spend three months experimenting with.
- Nausea, abdominal pain and rapid deep breathing alongside high glucose. That is an emergency.
Worth an ordinary appointment:
- Any value in the prediabetes range. It is the range where chapter 10 works best, which is the argument for acting on it rather than filing it.
- An HbA1c and a glucose that contradict each other.
- Prediabetes alongside anaemia or low ferritin, for chapter 5's reason.
- Diabetes in the family, previous gestational diabetes, or PCOS.
12. What we do not know
Whether intervening at prediabetes prevents hard outcomes. That it prevents and delays diabetes is established[8]. That fewer people consequently have heart attacks or live longer is a different and much harder question, and the evidence is weaker than the first sentence makes it sound.
What a fasting insulin means. No standardisation between laboratories, no guideline threshold, no trial steering by it. Everything in chapter 6 is mechanism and reasoning, which is less than the rest of this guide offers.
How long remission lasts. DiRECT reported 46% at one year[9]. What remains after ten, and in how many people, is not something one trial can answer.
Where the prediabetes line belongs. It is a convention, and the ADA and the WHO do not put it in the same place. A number just over an agreed line is exactly that.
And what a sensor means in a healthy person. No threshold has been validated against outcomes. Anyone telling you your spike after a banana is a problem does not know that.
13. Sources
- Emerging Risk Factors Collaboration; Sarwar N, Gao P, et al. Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet, 2010. PMID 20609967
- Selvin E, Steffes MW, Zhu H, et al. Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. New England Journal of Medicine, 2010. PMID 20200384
- American Diabetes Association Professional Practice Committee. 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes 2025. Diabetes Care, 2025. PMID 39651986
- Gonzalez A, Deng Y, Lane AN, et al. Impact of mismatches in HbA1c vs glucose values on the diagnostic classification of diabetes and prediabetes. Diabetic Medicine, 2020. PMID 31721287
- Madhu SV, Raj A, Gupta S, et al. Effect of iron deficiency anemia and iron supplementation on HbA1c levels: implications for diagnosis of prediabetes and diabetes mellitus in Asian Indians. Clinica Chimica Acta, 2017. PMID 27717800
- Li M, Zhang X, Zhou Y, et al. Interference of hemoglobin variants with HbA1c measurements by six commonly used HbA1c methods. Laboratory Medicine, 2024. PMID 38801245
- Matthews DR, Hosker JP, Rudenski AS, et al. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia, 1985. PMID 3899825
- Knowler WC, Barrett-Connor E, Fowler SE, et al. (Diabetes Prevention Program Research Group). Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. New England Journal of Medicine, 2002. PMID 11832527
- Lean ME, Leslie WS, Barnes AC, et al. Primary care-led weight management for remission of type 2 diabetes (DiRECT): an open-label, cluster-randomised trial. Lancet, 2018. PMID 29221645
- Umpierre D, Ribeiro PAB, Kramer CK, et al. Physical activity advice only or structured exercise training and association with HbA1c levels in type 2 diabetes: a systematic review and meta-analysis. JAMA, 2011. PMID 21540423
More free knowledge at kratosnatural.com.
© Kratos Natural. This guide is free: read it, keep it, print it, give it to whoever you like. What you may not do: sell it, or present it as your own work. The knowledge belongs to everyone. The text stays ours.