HbA1c is convenient: no fasting is needed, the pre-analytics are more stable, and it fluctuates less from day to day after yesterday's workout. That is why it is loved. And that is also why it is dangerous to love it blindly.
ADA 2026 Recommendation 2.4 (level B): in conditions where the relationship between HbA1c and glycaemia is disrupted, plasma glucose criteria are used for diagnosis. The source's list includes certain haemoglobin variants, pregnancy, glucose-6-phosphate dehydrogenase deficiency, HIV, and conditions that alter erythrocyte turnover.
One mechanism, different manifestations. HbA1c accumulates over the lifespan of the erythrocyte (about 120 days). Anything that shortens the cell's life pulls HbA1c down at the same mean glucose: haemolysis, blood loss, erythropoietin therapy, haemodialysis. Anything that ages the erythrocyte population pulls it up: iron deficiency anaemia. Haemoglobin variants (including sickle cell trait) interfere with some methods: ADA refers to the list of interferences at ngsp.org/interf.asp. The X-linked G6PD variant G202A, carried by about 11% of Black individuals in the US, is associated with a reduction in HbA1c of about 0.8% in homozygous men and 0.7% in homozygous women compared with people without the variant (ADA 2026, §2, citing study [18] of the section). This is not a "racial correction to the diagnosis". It is a genetic effect on the marker. ADA separately cautions against using race as a proxy for unstudied genetics.
The practical rule. If glucose values on the glucometer and in the diary are consistently 8–10 mmol/L while HbA1c is 5.8%, do not declare "excellent control". Look for haemolysis, transfusion, erythropoietin, a haemoglobin variant, laboratory interference. If it is the other way round — HbA1c 8.5% with fasting values of 5.5 — look for iron deficiency, a laboratory artefact, or postprandial peaks that the fasting snapshot did not catch. The alternative markers of chronic hyperglycaemia that ADA mentions for monitoring in case of discrepancy are fructosamine and glycated albumin. There are no fructosamine thresholds in Table 2.1 for the diagnosis of diabetes.
The X-linked glucose-6-phosphate dehydrogenase (G6PD) variant G202A was discussed above: it is a genetic effect on the marker, not a "racial correction". Point-of-care once more, this time as a quality trap. NGSP certification for monitoring ≠ clearance for diagnosis. Confusing "a method standardised to the DCCT" with "the DCCT study population" is a typical error in reading a lab report: the first is the calibration of the HbA1c assay, the second is an RCT in people with type 1 diabetes. In Kazakhstan, the 2022 protocol requires NGSP/DCCT and does not describe the glucometer as a diagnostic standard: "diagnosis is based on laboratory measurements of glucose levels".
Pregnancy is a separate universe of criteria that this course does not cover. ADA places gestational diabetes in Section 15. If the teaching case involves pregnancy, the correct move for the student is not to apply Table 2.1 "as for everyone", but to open the obstetric document.
The conclusion of the lesson. HbA1c is a powerful entry point in Table 2.1 as long as the erythrocyte behaves normally. As soon as cell turnover or haemoglobin changes, three doors remain, and all three are about plasma glucose.