Diagnostic Criteria for Type 2 Diabetes

Type 2 diabetes is diagnosed when any one of three standard blood tests crosses a specific threshold: a fasting plasma glucose at or above 126 mg/dL, a two-hour glucose reading at or above 200 mg/dL during an oral glucose tolerance test, or a hemoglobin A1c (HbA1c) at or above 6.5 percent. A fourth route exists for people already showing classic symptoms like excessive thirst, frequent urination, and unexplained weight loss: a random plasma glucose at or above 200 mg/dL, taken at any time of day. In practice, though, the story is messier than those clean cutoffs suggest, because the three main tests do not always agree on who has diabetes and who does not.

The Three Standard Blood Tests

Each of the three tests captures blood sugar from a different angle. Fasting plasma glucose (FPG) measures what your blood sugar does after an overnight fast, typically eight to twelve hours without food. The oral glucose tolerance test (OGTT) measures how your body handles a 75-gram glucose drink over two hours, essentially stress-testing your insulin response. And HbA1c reflects an average of your blood sugar over the past two to three months by measuring how much glucose has attached to your red blood cells’ hemoglobin.

The current diagnostic cutoffs are:

Fasting plasma glucose: 126 mg/dL (7.0 mmol/L) or higher on two separate occasions.
Two-hour OGTT: 200 mg/dL (11.1 mmol/L) or higher after drinking the glucose solution.
HbA1c: 6.5 percent or higher, confirmed on a repeat test unless symptoms are already present.
Random plasma glucose: 200 mg/dL or higher in the presence of classic hyperglycemic symptoms.

The fasting glucose threshold was lowered in the late 1990s from the older standard of 140 mg/dL down to 126 mg/dL, a change that brought the diagnosis more in line with the level of glucose at which complications like eye damage begin to appear. That shift was not without controversy. A 1999 study in JAMA found that roughly 60 percent of people newly classified as diabetic under the lower 126 mg/dL cutoff actually had normal HbA1c levels, raising questions about whether the threshold was catching people who did not yet have the chronic glucose elevations that drive real organ damage.¹ That debate has cooled over the years but never fully resolved, and it hints at a broader tension: the cutoffs are somewhat arbitrary lines drawn through a continuous biological spectrum.

Why the Tests Do Not Always Agree

If all three tests measured the same thing, they would produce the same verdict. They do not. FPG captures fasting glucose at a single moment, the OGTT captures how your body clears a glucose load, and HbA1c averages glucose exposure over months. Because they look at different windows, they regularly disagree on whether a given person is normal, prediabetic, or diabetic.

A study of a Malaysian cohort found that HbA1c flagged more people as having diabetes or prediabetes than fasting glucose did. The prevalence of diabetes was about 5.7 percent by FPG but 7.5 percent by HbA1c, and prediabetes rates climbed from about 10.6 percent to 14.2 percent when using HbA1c instead of fasting glucose. Yet HbA1c also had relatively low sensitivity for catching cases identified by glucose-based tests, meaning it was good at confirming diabetes when present but also missed a fair number of people who would have been diagnosed by glucose measures alone.²

The discordance can be dramatic. In a study of Filipino-American adults, five participants met the HbA1c criterion for diabetes, but none of those five had a fasting glucose above the diabetes threshold, and only one had a two-hour OGTT value in the diabetic range. Two of the five had completely normal fasting glucose and OGTT results despite an HbA1c of 6.5 percent or higher. For prediabetes, about two-thirds of those flagged met only the HbA1c criterion and would have been missed entirely by glucose testing alone.³

A large analysis using NHANES data put numbers on the problem from the other direction. When fasting glucose and two-hour OGTT were used together as the reference standard, the HbA1c cutoff of 6.5 percent missed about 75 percent of people with diabetes. For prediabetes, the 5.7 percent HbA1c threshold missed about 65 percent. Combining HbA1c with fasting glucose improved things for prediabetes detection but still left a sizable gap for diabetes itself.⁴ The takeaway for you: if one test comes back borderline or normal but your doctor still suspects a problem, a second type of test is worth doing.

The OGTT and What It Catches That Fasting Glucose Misses

The OGTT is the most burdensome test for patients: you have to fast overnight, drink a syrupy glucose solution, then sit for two hours while blood is drawn. Unsurprisingly, it is used less often than a simple fasting blood draw. But it catches something the other tests can miss: people who handle fasting just fine but cannot properly clear a glucose load. That distinction matters because impaired glucose tolerance is one of the strongest predictors of progression to full-blown diabetes and of cardiovascular risk.

A review in the European Journal of Internal Medicine noted that the OGTT detects early diabetes and impaired glucose tolerance more efficiently than fasting glucose, because the two-hour glucose value captures how your insulin response performs under challenge.⁵ More recently, researchers have looked at the one-hour mark during the OGTT as a potentially more sensitive diagnostic point. A study of over 1,800 patients compared one-hour glucose and two-hour glucose classifications. Using the one-hour reading, about 7.5 percent of patients were classified as diabetic and 29.8 percent as having impaired tolerance. Using the traditional two-hour reading, those numbers dropped to 5.6 percent and 18.1 percent. Around 16 percent of patients who looked normal at the two-hour mark already showed impaired tolerance at one hour.⁶ The shape of the glucose curve also mattered: patients with normal tolerance tended to have a biphasic curve (glucose rises, dips, and may rise slightly again), while those with impaired tolerance or diabetes more often had a monophasic curve that rose and fell without a secondary dip.

Some researchers have argued the two-hour OGTT threshold of 200 mg/dL is too low and should be raised to around 240 mg/dL, based on evidence that the lower cutoff catches people whose HbA1c is still normal and who may not develop the microvascular complications that define clinical diabetes.⁷ This remains a minority position, but it highlights the ongoing tension about where to draw diagnostic lines on a biological continuum.

The Prediabetes Zone

Before you hit the diabetes thresholds, there is a middle ground. Prediabetes is defined as fasting glucose between 100 and 125 mg/dL (called impaired fasting glucose, or IFG) or a two-hour OGTT value between 140 and 199 mg/dL (called impaired glucose tolerance, or IGT). For HbA1c, the prediabetes range is 5.7 to 6.4 percent. These categories were created to identify people at elevated risk for progressing to type 2 diabetes.⁸

Prediabetes is not destiny. A meaningful proportion of people in this range will revert to normal glucose levels, particularly if they lose weight, become more physically active, or both. But the risk is real: without intervention, roughly a quarter to a third of people with prediabetes will develop type 2 diabetes within five years. Identifying prediabetes is the primary reason screening programs exist, because lifestyle changes at this stage are genuinely effective at preventing or delaying the full disease.

Here, too, the tests disagree. A person can have impaired fasting glucose but normal glucose tolerance, or vice versa, and many people flagged by HbA1c alone have normal glucose-based tests. Because IFG and IGT identify somewhat different metabolic problems, having both at once carries the highest risk of progressing to diabetes.

When HbA1c Gives the Wrong Answer

HbA1c has become the most convenient diagnostic tool because it does not require fasting and reflects a longer window than a single glucose reading. But it relies on a key assumption: that your red blood cells live a normal lifespan (roughly 120 days) and that hemoglobin glycation happens at a standard rate. When either assumption fails, HbA1c can be misleadingly high or low.

Conditions that shorten red blood cell lifespan, such as hemolytic anemia, lead to falsely low HbA1c because the cells do not circulate long enough to accumulate glucose. Conversely, iron deficiency anemia can falsely raise HbA1c, because younger red blood cells accumulate glycation more aggressively and the cells themselves live longer on average. Hemoglobinopathies like sickle cell disease and thalassemia throw yet another wrench by altering how the hemoglobin molecule interacts with glucose, pushing the reading either direction depending on the specific variant and the assay used.⁹ In populations where these conditions are common, such as South Asia and sub-Saharan Africa, HbA1c-based diagnosis can be unreliable enough to warrant relying on glucose-based testing instead.

Race and ethnicity add another layer. At any given average blood glucose level, Black individuals tend to have HbA1c values about 0.4 percentage points higher than white individuals.¹⁰ The reasons are not fully understood, but may involve differences in hemoglobin glycation rates rather than actual glucose levels.¹¹ This means an HbA1c of 6.5 percent may represent a lower average glucose in a Black person than in a white person, potentially leading to overdiagnosis in Black populations using a single universal cutoff. One study found that the prevalence of microvascular complications like retinopathy was higher in African-Americans below the 6.5 percent threshold than in other groups, suggesting that the mismatch between HbA1c and actual glucose may have clinical consequences.¹² Whether the 6.5 percent cutoff should differ across racial groups remains an open question, though some researchers have argued for race-specific thresholds.¹³

Other conditions that affect HbA1c reliability include chronic kidney disease (which alters red cell turnover), recent blood transfusions, erythropoietin therapy, and pregnancy. If you have any of these, your doctor should be diagnosing diabetes based on glucose measurements rather than HbA1c.

Risk Starts Below the Diagnostic Threshold

One common misconception is that a fasting glucose below 100 mg/dL means you are completely in the clear. A large study tracking over 46,000 adults whose fasting glucose was entirely in the “normal” range found that diabetes risk climbed steadily with each additional milligram per deciliter. People with fasting glucose between 95 and 99 mg/dL were more than twice as likely to develop diabetes compared to those below 85 mg/dL, even after controlling for other risk factors like weight and age. Each single-point increase in fasting glucose raised diabetes risk by about 6 percent.¹⁴ That means a fasting glucose of 98 is not the same as 82, even though both are classified as normal. The current cutoffs tell you when you have crossed a clinical threshold, not when your risk starts.

Pregnancy and Early Gestational Diabetes

Pregnancy has its own set of diagnostic criteria, which are distinct from the thresholds for type 2 diabetes. If a woman’s fasting glucose at her first prenatal visit hits 126 mg/dL or above, she is considered to have preexisting (not gestational) diabetes. Between 92 and 125 mg/dL, the picture gets more complicated. A range of 5.1 to 6.9 mmol/L (roughly 92 to 125 mg/dL) during early pregnancy is now flagged as early gestational diabetes, a category that has emerged as clinicians have begun routinely screening fasting glucose at the first prenatal visit.¹⁵ Later in pregnancy, the standard gestational diabetes screening occurs at 24 to 28 weeks using a glucose challenge test, with its own set of cutoffs. HbA1c is generally not recommended for diagnosing gestational diabetes because the physiological changes of pregnancy (increased blood volume, altered red cell turnover) make it unreliable.

When It Is Not Actually Type 2

Crossing a diagnostic glucose or HbA1c threshold tells you that diabetes is present, but it does not tell you which type. Most adults diagnosed with diabetes have type 2, so clinicians default to that classification unless something looks unusual. But other forms of diabetes can masquerade as type 2, and getting the type wrong changes treatment fundamentally.

Latent autoimmune diabetes of adults (LADA) is an autoimmune form that appears in adulthood and can look identical to type 2 in the early years: the patient is often normal weight or slightly overweight, responds initially to oral medications, and has glucose levels in the same ranges. Over time, though, LADA progresses to complete insulin dependence because the immune system is destroying insulin-producing beta cells. C-peptide testing, which measures how much insulin your pancreas is producing, can help distinguish the two. People with LADA tend to have significantly lower C-peptide levels than those with type 2, and the diagnosis can be confirmed by testing for autoantibodies like anti-GAD.¹⁶

Maturity-onset diabetes of the young (MODY) is another imposter. It is a monogenic form of diabetes, meaning it is caused by a single gene mutation, and it typically appears before age 25. MODY accounts for a small percentage of all diabetes cases but is significantly underdiagnosed because most patients are managed as if they have type 1 or type 2 without further investigation. The standard glucose and HbA1c criteria will identify that diabetes is present, but they will not point toward MODY. Suspicion usually arises when a young, non-obese person develops diabetes with a strong family history spanning multiple generations and lacks the autoantibodies seen in type 1.¹⁷ Genetic testing confirms the diagnosis, and the treatment differs markedly depending on which gene is affected: some forms of MODY respond to low-dose sulfonylureas alone and never need insulin.

Screening Before the Blood Draw

Not everyone needs to rush to a lab. Risk-scoring questionnaires can help identify who should be screened in the first place, saving time and resources. The Finnish Diabetes Risk Score (FINDRISC) is one of the most widely validated. It uses eight simple questions about age, body mass index, waist circumference, physical activity, diet, blood pressure medication, history of high glucose, and family history of diabetes. A score above 10 on the FINDRISC has been shown to be a useful threshold for deciding who should get laboratory screening for undetected diabetes and prediabetes.¹⁸ Tools like these are particularly valuable in resource-limited settings where universal blood testing is not feasible, but they also serve as a reasonable first step for anyone wondering whether a diabetes test is warranted.

Continuous Glucose Monitors and Estimated HbA1c

Continuous glucose monitors (CGMs) are increasingly common, and they generate a metric called the glucose management indicator (GMI), which estimates what your HbA1c would be based on the sensor’s glucose readings. The appeal is obvious: if a device on your arm is already tracking glucose around the clock, why not use it for diagnosis? An observational study comparing CGM-derived estimates to lab-measured HbA1c found strong correlation and about 93 to 97 percent agreement within clinical limits, depending on the specific metric used.¹⁹ Still, CGM-derived metrics are not currently accepted as standalone diagnostic criteria. The same confounders that affect lab HbA1c (red cell turnover, hemoglobinopathies) are sidestepped by CGM since it measures glucose directly, which is actually an advantage. But sensor accuracy, calibration variability, and the lack of large-scale validation studies against clinical outcomes mean that CGM data is not yet ready to replace a lab draw for formal diagnosis. For people already wearing a CGM who see persistently elevated glucose readings, though, the data is a clear prompt to get proper diagnostic testing done.

How Diagnostic Thresholds Were Chosen

The specific numbers used as diagnostic cutoffs were not pulled from thin air, but they are not as biologically crisp as they might appear. The original logic behind the thresholds, dating back to epidemiological work in the 1970s and 1980s, was to find the glucose level above which the risk of diabetic retinopathy (damage to the blood vessels in the back of the eye) rises sharply. Population studies showed a roughly continuous rise in retinopathy risk with increasing glucose, but the curve steepened around a fasting glucose of 126 mg/dL and a two-hour OGTT value of 200 mg/dL. These inflection points became the diagnostic thresholds.

When HbA1c was adopted as a diagnostic criterion in 2010, the same logic applied: 6.5 percent was the level above which retinopathy risk appeared to climb more steeply. A systematic review and meta-analysis confirmed that the prevalence of both retinopathy and nephropathy was higher above that cutoff, supporting its specificity for diagnosis.²⁰ But as we saw, complications can appear below 6.5 percent in certain populations, and the relationship between glucose, glycation, and organ damage is not identical across all groups. The cutoffs work reasonably well as population-level diagnostic tools, but they are consensus decisions informed by data, not bright biological lines that separate “healthy” from “diseased.” For any individual, a number just below a threshold is not meaningfully different from a number just above it.