Does a normal A1C change with age or sex? What the ADA uses and what the data show

Search for a normal A1C by age and you will find tables with a separate band for every decade. None of them comes from the body that sets the diagnostic criteria. The American Diabetes Association Standards of Care 2026 print one set of A1C thresholds for nonpregnant individuals, and the section says the diagnostic and prediabetes tables are the criteria "that apply to children, adolescents, and adults." There is no adjustment for age and none for sex.

Two details in the same section shape how a borderline result should be read. The ADA notes that for all three tests (A1C, fasting glucose and the glucose tolerance test), risk is continuous, extending below the lower limit of the range and becoming disproportionately greater at the higher end. And outside a clear clinical diagnosis, a diagnosis needs confirming: two abnormal results, either at the same time, such as an A1C and a fasting glucose drawn together, or at two time points. When two different tests disagree, the ADA says the result above the cut point should be repeated, with careful consideration of factors that may affect measured A1C or glucose levels. The ADA does not list age among those factors. The authors of one study below argue it should be considered.

The cleanest evidence comes from studies that took people without diabetes, measured their glucose directly, and asked whether A1C still rose with age once glucose was accounted for. It does.

• Framingham and NHANES (Pani and colleagues, Diabetes Care 2008). In 2,473 Framingham Offspring participants and 3,270 NHANES 2001 to 2004 participants without diabetes, A1C rose 0.014 and 0.010 percentage points per year (0.012 in the Framingham participants with normal glucose tolerance). The rise held in the Framingham participants with normal glucose tolerance, and after adjusting for sex, BMI, fasting glucose and two-hour glucose.
• NHANES III (Davidson and Schriger, 2010). In 2,712 adults aged 40 to 74 without a diabetes history, A1C rose 0.10 percentage points per decade in people with normal glucose tolerance, independent of their fasting and two-hour glucose.
• SIGT and NHANES 2005 to 2006 (Dubowitz and colleagues, 2014). In people with normal glucose tolerance, A1C rose 0.07 percentage points per 10 years. The authors calculated that an 80-year-old would carry an A1C 0.35 points higher than a 30-year-old with the same normal glucose tolerance.

Read that table for what it is. It describes the A1C values a national sample of Americans without diagnosed diabetes actually had in the early 2000s, and about a third of them (31.6 percent) had impaired fasting glucose, so it is not a picture of perfect metabolic health. It is not a set of normal ranges, and no guideline uses it as one. Its value is the shape: between the under-40s and the over-70s the mean rises by about three tenths of a point and the 97.5th percentile by about half a point. (The paper's abstract gives 5.6 percent for the under-40 97.5th percentile; its Table 1, reproduced here, gives 5.7.)

Pani and colleagues found that the age effect survived adjustment for glucose, which points to something other than blood sugar contributing. Dubowitz and colleagues drew the practical consequence: the specificity of A1C-based criteria for prediabetes decreased substantially with increasing age, meaning more older adults are flagged by A1C who would not be flagged by a glucose test. Davidson and Schriger reported the same pattern at the diabetes line. Among adults aged 40 to 74 with an A1C of 6.5 to 6.9 percent, only 65 percent met fasting or oral glucose tolerance criteria for diabetes. In the one age comparison their abstract reports, among non-Hispanic Black adults, 68 percent of those aged 40 to 74 and 87 percent of those over 64 would not have had diabetes by those criteria.

None of these authors argued that a rising A1C is harmless, and the ADA has not adopted age-adjusted cut points. Their own conclusion is narrower: Dubowitz and colleagues wrote that age should be taken into consideration when using A1C to diagnose diabetes and prediabetes. For a result in the diabetes range, the ADA already requires a second abnormal result, from a repeat or from a different test such as a fasting plasma glucose, unless the diagnosis is clinically clear.

The ADA uses one threshold for women and men. The measured values differ slightly, and the difference depends on age.

• A 1997 study of 4,580 healthy adults in Taiwan found men had higher average glycohemoglobin than women under 55, and the authors described a lower average level in women before menopause.
• A 2025 analysis of 17,271 same-day A1C and fasting glucose pairs from two English hospital laboratories found median A1C 1 mmol/mol lower in women under 45 than in men (34 against 35 mmol/mol), and no difference at 45 and over. These were people whose doctor ordered the test, not a population sample.
• In the Framingham and NHANES analysis, women in the peri- and postmenopausal age groups had a steeper rise in A1C with age than men. The authors suggested, without proving it, that lower hemoglobin and faster red cell turnover in menstruating women may explain part of the earlier gap.

To put 1 mmol/mol in familiar units: the ADA pairs 5.7 percent with 39 mmol/mol and 6.5 percent with 48 mmol/mol, so by our arithmetic one mmol/mol is about 0.09 of a percentage point. A real difference, and a small one.

Risk may not track identically either. In 608,474 Alberta adults aged 40 to 79 without diabetes or heart disease, an A1C of 5.5 to 5.9 percent carried a 12 percent higher adjusted risk of cardiovascular hospitalization in men than an A1C of 5.0 to 5.4, and no measurable increase in women. At 6.0 to 6.4 percent the risk rose in both, by 38 percent in men and 17 percent in women. No guideline turns that into a sex-specific normal range.

The ADA addresses this directly, and calls racial differences in A1C controversial. Its cut points carry no race adjustment. Studies have found that Black individuals have slightly higher A1C levels, approximately 0.3 percentage points, than non-Hispanic White or Hispanic people, and a G6PD variant carried by 11 percent of Black individuals in the US lowers A1C by about 0.8 percentage points in homozygous men and 0.7 in homozygous women. The same section warns against using race or ancestry as proxies for poorly understood genetic differences, and notes that the association of A1C with complications appears similar in Black and non-Hispanic White populations. A continuous glucose monitoring study of people with type 1 diabetes found A1C ran 0.4 points higher in Black than in White participants at the same mean glucose; the authors wrote that race only partially explained the gap, and too few participants had an A1C under 6.5 percent to apply the finding to the normal range. Our guide to what your A1C means covers the hemoglobin variants and conditions that distort the assay.

Much of the "A1C by age" confusion comes from mixing two different numbers. The diagnostic thresholds above define who has diabetes. Treatment goals say how far to lower A1C in someone who already does, and those are individualized, with age and health status as explicit inputs. The ADA's general goal is an A1C under 7 percent for many nonpregnant adults, with lower goals such as under 6.5 percent possible for people in good health with low treatment risks. For older adults, Section 13 sets out a framework.

The reasoning is about treatment harm, not about what is normal. The ADA states that people with shorter life expectancy, frailty or substantial cognitive or functional impairment will have less benefit from glucose lowering and are more likely to experience serious adverse effects such as hypoglycemia. So an A1C of 7.8 percent can sit inside the suggested goal for an older adult in the complex-health group and still be in the diabetes range for anyone being screened.

Our A1C reference page carries the ADA thresholds with their source and effective date, and each band has its own page: normal, prediabetes, diabetes and the treatment goal for many nonpregnant adults. Put your own number into the lab result lookup to see which band its source puts it in, or convert it with the A1C to average glucose converter.

If you have never been tested, the ADA says screening should begin at age 35 for everyone, earlier with overweight and risk factors, and be repeated at a minimum of three-year intervals when normal. Our type 2 diabetes screening page sets out who should be screened and how often, and the metabolic numbers to know after 40 puts A1C beside the other markers that move in midlife.

Bring any A1C of 5.7 or above to a review, and, if an A1C near a threshold is your only abnormal result, ask whether a fasting glucose would help confirm it, especially after 60. If you already have diabetes and are older, ask which of the ADA health-status groups your clinician places you in and what A1C goal follows from it, and say so if you have had low blood sugar episodes, because the ADA ties less stringent goals to exactly that risk.