TL;DR: Glycated haemoglobin, HbA1c, measures the proportion of your haemoglobin that has sugar stuck to it, which gives an integrated picture of average blood glucose over roughly the preceding two to three months. It is the standard test for diagnosing and monitoring diabetes, and it is also a decent general marker of metabolic health. The relationship with mortality is not a straight line: it is J-shaped or U-shaped in most large cohorts, meaning risk rises as HbA1c climbs into the diabetic range, but also rises at the very low end, largely because low values often reflect anaemia, liver disease, or frailty rather than superb glucose control. Diabetes substantially shortens life expectancy, and that is well quantified. What is less intuitive is that in people who already have type 2 diabetes, driving HbA1c very low with medication has not reliably reduced mortality in trials, and one large trial found the opposite. The tissue damage of hyperglycaemia is real. The idea that lower is always better is not supported. Exercise, particularly resistance training, and fat loss remain the interventions with the best combination of efficacy and safety.
What does HbA1c actually measure?
Glucose in your blood attaches non-enzymatically to proteins, including the haemoglobin inside your red blood cells. This process, glycation, happens continuously and at a rate proportional to how much glucose is around. Red blood cells live roughly 120 days, so at any moment your circulating red cells carry a record of their glucose exposure across their lifespan. HbA1c reports the percentage of haemoglobin that has been glycated in this way.
Because older cells have had more exposure than younger ones, the measure is weighted towards the recent past. Roughly half the value reflects the preceding month, with the remainder spread across the two months before that. This is why HbA1c responds to a change in glucose control within about six to twelve weeks rather than immediately, and why testing it more often than every three months is usually pointless.
The units are a small source of confusion. The older DCCT units express HbA1c as a percentage, which is what most people know. The IFCC standard expresses it in mmol/mol, which is what UK laboratories typically report. They are the same measurement in different clothing: 5.7% corresponds to about 39 mmol/mol, 6.5% to about 48 mmol/mol, 7.0% to about 53 mmol/mol.
The key conceptual point is that HbA1c is an average, and averages hide things. Two people with identical HbA1c can have quite different glucose patterns, one with a stable line and one with large post-meal excursions balanced by lower values elsewhere. Whether that variability matters independently is an open question that continuous glucose monitoring has made easier to ask than to answer.
What HbA1c level is associated with the lowest mortality?
Broadly, the range around 5.0 to 5.6 percent, roughly 31 to 38 mmol/mol, though the exact nadir differs between cohorts and depends heavily on who is in them.
The shape matters more than the number. In most large cohort studies, including analyses from NHANES and from UK-based primary care data, the relationship between HbA1c and mortality is J-shaped or U-shaped rather than linear. Risk climbs as HbA1c rises into the prediabetic and diabetic ranges, which is unsurprising. But risk also rises at the bottom end, and that part surprises people.
The low-end rise is mostly not a story about glucose being too low. It is a story about what causes a low HbA1c. Anaemia and conditions with shortened red cell survival lower HbA1c mechanically, because cells do not live long enough to accumulate glycation. Chronic liver disease, some blood disorders, and recent blood loss do the same. Frailty, unintentional weight loss, and occult illness are also associated with lower values. So the people at the bottom of the HbA1c distribution include a substantial number who are unwell for reasons unrelated to glucose, and their mortality risk reflects that illness, not their glucose. This phenomenon, where reverse causation creates a spurious association at one end of a distribution, is a recurring trap in epidemiology and it appears with cholesterol and body weight too.
The standard diagnostic thresholds, used by the WHO, the ADA, and NICE with some variation: below 5.7% (39 mmol/mol) is normal, 5.7 to 6.4% (39 to 47 mmol/mol) is prediabetes or non-diabetic hyperglycaemia, and 6.5% (48 mmol/mol) or above on two occasions supports a diagnosis of diabetes. These are administrative cut points on a continuous variable, chosen partly for their relationship with retinopathy risk. Nothing biological changes at 6.5%.
Does prediabetes actually shorten your life?
The association is real, and the framing deserves care.
Prediabetes is associated with higher cardiovascular risk and higher all-cause mortality than normal glycaemia in cohort studies, though the increment is modest compared with the jump seen in established diabetes. It is also strongly associated with the surrounding metabolic picture, adiposity, blood pressure, triglycerides, so isolating the contribution of the HbA1c value itself is difficult.
More importantly, prediabetes is a risk state, not a disease, and a large proportion of people with prediabetes never progress to diabetes. Some revert to normal glycaemia spontaneously. This is why some clinicians are uncomfortable with the label: it converts a large fraction of the adult population into patients, with attendant anxiety, without a clear benefit for many of them.
Where the evidence is genuinely strong and genuinely encouraging is intervention. The Diabetes Prevention Program in the United States and the Finnish Diabetes Prevention Study both randomised people with impaired glucose tolerance to intensive lifestyle intervention, focused on modest weight loss and regular physical activity, versus control. Both found substantial reductions in progression to type 2 diabetes. In the DPP, the lifestyle arm outperformed metformin. Long-term follow-up of these cohorts has shown the benefit persists for years after the active intervention ends. This is one of the better demonstrations in medicine that a behavioural intervention can change a disease trajectory.
So prediabetes is best understood as useful information rather than a sentence. It is a signal that the metabolic trend is in the wrong direction, at a point when it is comparatively easy to change.
How much life does type 2 diabetes cost?
This is where the numbers get serious, and honest reporting requires acknowledging they are averages that hide enormous variation.
Large pooled analyses of cohort data, including work from the Emerging Risk Factors Collaboration, have estimated that diabetes is associated with a reduction in life expectancy on the order of six years for someone diagnosed at around age 50, with a wider gap for earlier diagnosis. Type 2 diabetes diagnosed in youth or early adulthood carries a considerably larger burden, because the exposure time is longer and because early-onset disease tends to be more aggressive.
The mechanism is not mysterious. Chronic hyperglycaemia damages small and large vessels. Microvascular consequences include retinopathy, nephropathy, and neuropathy. Macrovascular consequences are coronary disease, stroke, and peripheral arterial disease, and cardiovascular disease remains the leading cause of death in people with diabetes. Diabetes is also associated with higher rates of several cancers, of dementia, and of infection.
But these figures come from cohorts recruited under earlier treatment paradigms. Diabetes care has improved substantially, particularly with drug classes that have demonstrated cardiovascular and renal outcome benefits in trials, and the life expectancy gap in a person diagnosed today with good care is likely narrower than the historical estimates suggest. Well-controlled diabetes with attention to blood pressure and lipids is a very different prognosis from poorly controlled diabetes.
Does lowering HbA1c with drugs make you live longer?
This is the most counterintuitive part of the topic and the part most often glossed over.
You would expect that if high glucose causes damage, lowering glucose more aggressively would reduce that damage proportionally. The trials do not straightforwardly support that.
ACCORD randomised people with type 2 diabetes at high cardiovascular risk to intensive glucose lowering, targeting an HbA1c below 6.0%, versus standard treatment. The intensive arm was stopped early because of increased mortality. That result was genuinely shocking to the field and has never been fully explained, though hypoglycaemia and the drug combinations used to achieve the target are the usual suspects. ADVANCE and VADT, two other large trials of intensive glucose lowering, did not find mortality benefit either, though they did not find the ACCORD harm signal.
What intensive glucose control does reliably deliver is reduced microvascular complications: less retinopathy, less nephropathy. That is worth having. The UKPDS, which followed people with newly diagnosed type 2 diabetes over many years, found microvascular benefit from tighter control and, in long-term follow-up, a legacy effect on cardiovascular outcomes from early good control. The timing appears to matter: good control early in the disease seems to pay off later, while aggressive control imposed late in the disease on people who already have vascular damage does not.
The other thing that changed the field: several newer drug classes, notably SGLT2 inhibitors and GLP-1 receptor agonists, have demonstrated cardiovascular and mortality benefits in outcome trials, and these benefits appear larger than would be predicted from their glucose-lowering alone. That reinforces the interpretation that the drug matters, not just the number it produces. Lowering HbA1c is not a mechanism. It is a readout.
The practical implication is that chasing a low HbA1c number for its own sake is a mistake, and that individualised targets, higher for older or frailer people at risk from hypoglycaemia, lower for younger people early in the disease, are the standard of care for good reasons.
What actually improves HbA1c?
The interventions are unglamorous and effective.
Losing excess body fat is the most powerful single lever in type 2 diabetes. Trial evidence, including the DiRECT study in UK primary care, has shown that substantial weight loss achieved through structured dietary intervention can put type 2 diabetes into remission in a meaningful proportion of people, particularly those diagnosed within the last few years. That finding reframed type 2 diabetes from an inevitably progressive condition to one that is, for some, reversible.
Physical activity works through several routes. Muscle contraction moves glucose into cells through a pathway that does not require insulin, which is why a walk after a meal blunts the glucose rise. Regular training improves insulin sensitivity independent of weight loss. Resistance training deserves particular mention because skeletal muscle is the largest site of glucose disposal in the body, and more muscle means more capacity to handle a glucose load. Our strength training and mortality post covers the wider case, and the exercise for longevity protocol sets out how to structure it. Even breaking up prolonged sitting helps, as discussed in sitting time and mortality, and the walking evidence is relevant to the simplest version of this.
On diet, the honest summary is that no single dietary pattern has been shown to be uniquely superior, and adherence predicts outcomes better than macronutrient composition. Low-carbohydrate approaches lower HbA1c effectively in the short to medium term, which is unsurprising given the direct relationship, though long-term adherence is often poor and long-term mortality data are lacking, as covered in keto and longevity. Mediterranean and DASH-style patterns have good evidence for cardiovascular outcomes in people with metabolic risk, discussed in DASH diet and lifespan. Reducing refined carbohydrate and ultra-processed food intake is common to nearly every approach that works.
Sleep matters more than people expect. Experimental sleep restriction reliably impairs insulin sensitivity in healthy volunteers within days. It is a real physiological effect, not a soft factor.
When is HbA1c the wrong test?
Several situations make HbA1c unreliable, and they are common enough to matter.
Anything that changes red cell lifespan distorts the result. Haemolytic anaemia, sickle cell disease and trait, thalassaemia, recent blood transfusion, recent significant blood loss, and iron deficiency anaemia all shift HbA1c, sometimes substantially, in directions that have nothing to do with glucose. Chronic kidney disease and advanced liver disease also interfere. Pregnancy alters red cell turnover, which is why gestational diabetes is diagnosed with glucose tolerance testing rather than HbA1c.
Some haemoglobin variants interfere with certain assay methods, producing spurious results. Laboratories are generally aware of this, but it is worth knowing if your result seems wildly inconsistent with your clinical picture.
In these situations, fasting glucose, an oral glucose tolerance test, or fructosamine may be used instead. If your HbA1c does not match what you would expect, the possibility that the test is wrong is worth considering before assuming your metabolism is.
Frequently Asked Questions
Is a lower HbA1c always better? No, and this is one of the more important misconceptions to correct. The relationship with mortality is J-shaped, so risk rises at both ends. At the low end that is largely because low values often reflect anaemia, liver disease, or underlying illness rather than excellent glucose control. Separately, in people with diabetes, driving HbA1c very low with medication has not reduced mortality in trials, and the ACCORD trial found increased mortality in the intensive arm. Targets should be individualised with a clinician.
Can I reverse prediabetes? Frequently, yes. The Diabetes Prevention Program and the Finnish Diabetes Prevention Study both showed that intensive lifestyle intervention focused on modest weight loss and regular physical activity substantially reduced progression to type 2 diabetes, with benefits persisting for years afterwards. Many people with prediabetes revert to normal glycaemia. Prediabetes is best treated as an early warning at a point when change is comparatively easy, not as a diagnosis of an irreversible condition.
How often should I test my HbA1c? Because HbA1c reflects roughly three months of glucose exposure, testing more often than every three months rarely adds information. For someone without diabetes and without risk factors, routine testing every few years as part of general health checks is typical. For someone with prediabetes, annual testing is common. For someone with diabetes, three to six monthly is usual depending on stability. Your clinician should set the interval based on your situation.
Should I use a continuous glucose monitor instead? CGMs give you a level of detail HbA1c cannot: post-meal excursions, overnight patterns, glycaemic variability. In type 1 diabetes and insulin-treated type 2 diabetes they have clear clinical value. In people without diabetes, the evidence that CGM data improves health outcomes is currently thin, and there is a real risk of anxiety about normal physiological glucose rises that occur in everyone after eating. It can be informative, but do not mistake more data for better decisions.
Does eating sugar cause a high HbA1c? Indirectly and partially. HbA1c reflects average blood glucose, which is driven by total carbohydrate intake, insulin sensitivity, body fat, physical activity, and genetics together. A high sugar intake contributes, particularly through sugar-sweetened drinks and their effect on body weight, but insulin resistance driven by excess adiposity and inactivity typically matters more than sugar intake per se. This is why fat loss and exercise usually move HbA1c more than cutting sugar alone.
Medical Disclaimer
This article is for general information only and does not constitute medical advice, diagnosis, or treatment. It is not a substitute for consultation with a qualified healthcare professional who knows your medical history. Biomarker results, including HbA1c, require clinical context to interpret and can be unreliable in the presence of anaemia, haemoglobin variants, kidney or liver disease, recent transfusion, and pregnancy. Do not start, stop, or change any medication or supplement on the basis of this content, and never adjust diabetes medication yourself, as hypoglycaemia can be dangerous. Do not act on a laboratory result without discussing it with a clinician who can interpret it alongside your full health picture. If you have symptoms that concern you, seek medical attention rather than relying on information found online.