When fasting glucose and HbA1c disagree: why the two numbers diverge and what is checked next

Fasting glucose and HbA1c can place the same person in different diagnostic categories, as US survey data illustrate (Cowie et al., 2010). A mismatch alone does not establish that either result is wrong. The tests reflect different aspects of glucose exposure (ADA, 2026). In US adults without diagnosed diabetes, fasting glucose varied more between repeat measurements than HbA1c did (Selvin et al., 2007). Red cell survival can also affect HbA1c, as a small human study illustrates (Cohen et al., 2008). Knowing why the two numbers diverge tells you which question is worth asking next. What each number means on its own is covered in our piece on fasting glucose and HbA1c.

Two tests, two different questions

Fasting plasma glucose is measured after at least eight hours without caloric intake (ADA, 2026). It is the glucose concentration at that one moment. HbA1c is the share of haemoglobin that has glucose attached to it. Because glucose binds to haemoglobin for as long as a red blood cell circulates, HbA1c reflects average blood glucose over roughly the previous two to three months (ADA, 2026).

For nonpregnant individuals, the American Diabetes Association uses the following cut-offs: a fasting glucose of 100–125 mg/dL (5.6–6.9 mmol/L) or an HbA1c of 5.7–6.4% counts as prediabetes; 126 mg/dL (7.0 mmol/L) or higher, or 6.5% or higher, is in the diabetes range (ADA, 2026).

Because the criteria describe different aspects of glucose control, they do not select the same people. In the NHANES 2005–2006 comparison among US adults aged 20 or older, HbA1c identified 30% of those classified as having undiagnosed diabetes by any of three criteria: HbA1c, fasting plasma glucose, or two-hour plasma glucose (Cowie et al., 2010). In that survey, the diagnostic groups overlapped but did not coincide.

A single fasting glucose moves more than it seems

In 685 US adults without diagnosed diabetes who had their tests repeated about two weeks apart (NHANES III), the within-person coefficients of variation were 5.7% for fasting glucose and 3.6% for HbA1c. Of the people whose first fasting glucose was 126 mg/dL or higher, about 70% were at that level again on the second test (Selvin et al., 2007).

Handling of the sample matters too, because blood cells continue consuming glucose after collection. In laboratory experiments on venous blood from nonfasting employee volunteers, mean glucose fell by 4.6% at 2 hours and 7.0% at 24 hours in tubes containing sodium fluoride and oxalate, compared with 0.3% and 1.2% in acidified tubes (Gambino et al., 2009).

This is one reason a single result is usually not enough. Confirming diabetes generally requires two abnormal results: different tests taken together or separately, or one test repeated. An exception is classic hyperglycaemic symptoms or crisis with a random plasma glucose of 200 mg/dL (11.1 mmol/L) or higher (ADA, 2026).

When HbA1c is not following glucose

Red cell survival and haemoglobin characteristics can complicate HbA1c interpretation. The studies below illustrate specific findings and their limits.

• Iron deficiency. A systematic review included 12 studies of non-pregnant adults not known to have diabetes. The studies generally reported higher HbA1c with iron deficiency, with or without anaemia, without a corresponding increase in glucose measures; the findings were not uniform. Other forms of anaemia may lower HbA1c, but the data there are limited (English et al., 2015).

• Haemoglobin variants. In a retrospective analysis of 4,620 African American adults from CARDIA and the Jackson Heart Study, sickle cell trait was associated with HbA1c about 0.3 percentage points lower at a given fasting glucose level. Among observations from participants with no prior diabetes diagnosis and no diabetes medication use, HbA1c-defined prediabetes was present in 29.2% with the trait and 48.6% without it (Lacy et al., 2017).

• Red cell survival in people with normal blood counts. In a small study of 12 haematologically normal participants, mean circulating red cell age ranged from 38 to 60 days in six people without diabetes and from 39 to 56 days in six with diabetes. The authors calculated that this variation could materially alter HbA1c at a given average glucose (Cohen et al., 2008).

ADA 2026 recommends plasma glucose criteria when the relationship between HbA1c and glycaemia is altered, including with certain haemoglobin variants, pregnancy, G6PD deficiency, HIV, or altered red cell turnover — for example haemodialysis, recent blood loss or transfusion, or erythropoietin therapy (ADA, 2026).

What is usually checked next

For discordant diagnostic results, the ADA advises repeating the test that is above the diabetes threshold and considering interference with either measurement (ADA, 2026). In practice, the questions worth taking to a doctor are:

1. Should the abnormal test be repeated on a properly handled sample?

2. If HbA1c looks out of line with glucose, is there a reason to check the blood count and iron status?

3. Could a haemoglobin variant interfere with the laboratory's HbA1c method?

4. Would measuring two-hour plasma glucose during a 75-g oral glucose tolerance test be appropriate (ADA, 2026)?

These are questions, not a plan: which of them apply depends on the rest of the picture, and that is a decision for the doctor.

Why the pair still matters

Discordance does not make either number unimportant. In ARIC, 11,092 middle-aged Black and White US adults without known diabetes or prior cardiovascular disease were followed for subsequent outcomes. HbA1c was associated with later diabetes and cardiovascular events; its association with all-cause mortality was J-shaped. These associations remained after adjustment for fasting glucose. In the fully adjusted categorical analysis, fasting glucose was not significantly associated with cardiovascular events or death after HbA1c was included, although its association with later diabetes remained (Selvin et al., 2010). That is an observational association in one cohort; it does not show that lowering HbA1c by itself changes those outcomes.

What this does not tell you

• A discordant pair is not proof that one test is wrong. Both can be accurate and describe different things.

• The studies above come from specific groups: US survey data, African American cohorts for sickle cell trait, and a 12-person study of red cell survival. These results should not be assumed to apply unchanged to other populations.

• Nothing here tells you which test is "better" for you personally.

In a personal record it helps to keep both numbers side by side, with the date and the laboratory. Bring earlier results to the discussion, noting any changes in collection conditions or laboratory method. Repeated results still need to be interpreted for possible interference. What a summary can and cannot tell you is covered separately.

This material is for information only and does not replace a consultation with a doctor.

References

• American Diabetes Association Professional Practice Committee for Diabetes (2026). 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes—2026. Diabetes Care, 49(Suppl 1), S27–S49. PMID 41358893.

• Cowie C.C. et al. (2010). Prevalence of diabetes and high risk for diabetes using A1C criteria in the U.S. population in 1988–2006. Diabetes Care, 33(3), 562–568. PMID 20067953.

• Selvin E. et al. (2007). Short-term variability in measures of glycemia and implications for the classification of diabetes. Archives of Internal Medicine, 167(14), 1545–1551. PMID 17646610.

• Gambino R. et al. (2009). Acidification of blood is superior to sodium fluoride alone as an inhibitor of glycolysis. Clinical Chemistry, 55(5), 1019–1021. PMID 19282354.

• English E. et al. (2015). The effect of anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic review. Diabetologia, 58(7), 1409–1421. PMID 25994072.

• Lacy M.E. et al. (2017). Association of sickle cell trait with hemoglobin A1c in African Americans. JAMA, 317(5), 507–515. PMID 28170479.

• Cohen R.M. et al. (2008). Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood, 112(10), 4284–4291. PMID 18694998.

• Selvin E. et al. (2010). Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. New England Journal of Medicine, 362(9), 800–811. PMID 20200384.