What an A1C Measures, and What a Percentage Point of It Is

Answer first: it counts what glucose left behind, not what is in the blood now

Hemoglobin is the part of a red blood cell that carries oxygen. Glucose circulating in the blood attaches to it, and it stays attached for the life of the cell. The higher the glucose level in the bloodstream, the more of it binds.

The A1C test measures the amount of hemoglobin with glucose attached, and reports it as a percentage of total hemoglobin. That is the whole quantity. It goes by several names — hemoglobin A1C, HbA1c, glycated hemoglobin, glycohemoglobin — and they all refer to the same measurement.

Because red cells live for months, the result reflects average blood glucose over roughly the past three months rather than the level at the moment of the draw. That is why the test does not require fasting and can be run at any time of day. It is also why it answers a different question from a finger stick, and why the two are not interchangeable.

The bands, and who certifies the test

The National Institute of Diabetes and Digestive and Kidney Diseases publishes the diagnostic ranges used in the United States: normal below 5.7 percent, prediabetes from 5.7 to 6.4 percent, and diabetes at 6.5 percent or above. Any test used to diagnose diabetes requires confirmation with a second measurement unless there are clear symptoms.

There is a standardization step behind those numbers that most people never hear about. For diagnosis, a blood sample drawn from a vein goes to a lab using a method certified by the NGSP, formerly called the National Glycohemoglobin Standardization Program. That body certifies that makers of A1C tests produce results consistent and comparable with the ones used in the Diabetes Control and Complications Trial — the trial the clinical meaning of these numbers was originally established in.

That certification is the reason an A1C from one lab means the same thing as an A1C from another. The institute also notes that samples analyzed in a doctor's office or clinic, known as point-of-care tests, should not be used for diagnosis. It adds that the test should not be used to diagnose type 1 diabetes, gestational diabetes, or cystic fibrosis-related diabetes.

A percentage point is not a percent

This is the single most common misreading of the number, and it appears constantly in coverage of these drugs.

The result is itself a percentage: the share of hemoglobin carrying glucose. So when a label reports a change of 0.3, that is 0.3 percentage points of that share — a move from, say, 5.8 to 5.5. It is not a 0.3 percent reduction of the number, which would be a change of about two hundredths of a point and would be invisible.

The distinction matters because the two readings differ by a factor of roughly twenty. A sentence that says a drug lowered A1C by 0.3 percent, when the source said 0.3 percentage points, has understated a real result almost to nothing. The reverse error, reading a percentage-point change as though it were a proportional one, overstates it.

Careful sources say points, or percentage points, precisely to keep those apart. When a page says only percent, the underlying table is usually the only way to tell which it meant.

The same drug moves it very differently depending on who is in the trial

The semaglutide weight label reports this measure in the same cardiometabolic tables that carry waist circumference and blood pressure, and reading down those tables shows something that no headline captures.

In the trials enrolling adults with obesity or overweight and no diabetes, baseline values sat close to normal — 5.4, 5.7, and 5.8 in different studies. The changes reported are correspondingly small: falls of 0.1 to 0.5, against placebo changes near zero, for differences from placebo of 0.2 to 0.3.

In the trials that enrolled people with type 2 diabetes, the same tables show baselines of 8.1 and 8.2, and treated groups falling 1.6 and 1.7 against placebo falls of 0.2 and 0.4.

Same molecule, same measurement, results five times larger in one population than the other. The reason is arithmetic rather than pharmacology: a value already near the normal range has very little room to fall, so a trial in that population cannot produce a large movement no matter what the drug does. This is why a blood-sugar result lifted out of a diabetes trial and set beside a weight trial is not a fair comparison, and why the FDA guidance lists A1C as a secondary endpoint specifically in subjects with type 2 diabetes.

Where it sits in a trial's statistics, and where it sits in the entry criteria

In the weight trials this measure is a secondary endpoint. The semaglutide label footnotes those tables with a line stating the listed parameters were not included in the prespecified hierarchical testing — the mechanism that protects a trial against false positives from testing many things — with named exceptions. In one of the earlier trials the blood-sugar row is inside the hierarchy; in one of the newer pair it is inside for a specific comparison. Elsewhere it is outside.

In the diabetes products it is the opposite: the primary endpoint, the thing the trial was built to move, with weight reported alongside it.

The measure also does duty as an entry rule. The tirzepatide cardiovascular outcomes trial required patients to have an A1C of at least 7.0 percent and no more than 10.5 percent to enroll, alongside established cardiovascular disease and an index of at least 25. The FDA guidance recommends that weight-reduction trials generally exclude subjects with poor control at baseline. Its example is an A1C above 10 percent, or fasting glucose above 270 milligrams per deciliter. It also recommends stratifying randomization by baseline A1C, with 8 percent as its illustrative cut point.

So the same number is, depending on where it appears, the thing a trial is measuring, a supporting observation, or a gate deciding who gets in at all.

What can make the number wrong

Because the test depends on red blood cells, anything that changes how long those cells live changes the result. The institute lists recent blood loss, sickle cell disease, erythropoietin treatment, hemodialysis, and transfusion as conditions that can change A1C levels. A falsely high result can occur in people who are very low in iron, such as those with iron-deficiency anemia. Kidney failure and liver disease are named as other causes of false results too.

The institute also notes the test is less affected by short-term changes than a fasting glucose or a tolerance test. That is its main advantage. It then states the limit of the same property: large changes in blood glucose over the past month will show up, but the test does not show sudden, temporary increases or decreases. And though the result represents a long-term average, levels within the past thirty days have a greater effect on the reading than earlier months.

There is one more conversion worth knowing about, because labs sometimes print it. Estimated average glucose is calculated from the A1C and converts the percentage into the milligrams-per-deciliter units a home meter uses. The institute is explicit that it will not match daily readings, because it is a long-term average rather than a level at a single moment.

How to read a blood-sugar claim on a program page

Four questions, and they are the same ones that work on a weight percentage.

In whom. A result from a trial in people with type 2 diabetes and a result from a weight trial in people without it are different measurements of different populations, and the first will always be larger.

From what baseline. A group starting near the normal range has almost no room to move, so a small change there is not evidence of a weak drug and a large change in a group starting high is not evidence of a strong one.

In what unit. Points and percent differ by roughly a factor of twenty, and only the underlying table settles which one a sentence meant.

And against what. Placebo groups in these trials were also enrolled in a diet and exercise program, and their blood-sugar rows moved too. The difference from placebo is the figure that isolates the medication, and it is printed beside the raw change on every one of these tables.