Normal Blood Sugar Is Not Remission

In a two-year trial, 60.2% of people taking tirzepatide reached an HbA1c below 5.7% — the non-diabetic range — against 24.0% of those receiving intensified conventional care.^[1] That result is going to be described as reversing or curing diabetes. It is neither, for a reason that is not a technicality, and understanding why is the difference between a realistic plan and a painful surprise.

The comparator is what makes this trial unusual

Most trials in this field compare a drug against placebo, which answers a question nobody faces in a clinic. SURPASS-EARLY did something harder. It enrolled 794 adults across 78 sites in ten countries, all within four years of a type 2 diabetes diagnosis and all inadequately controlled on metformin, and randomized them to tirzepatide at 15 mg or the maximum they could tolerate, or to intensified conventional care — whatever their clinicians would otherwise do, guided by local treatment guidelines.

That comparison arm was allowed to include other GLP-1 receptor agonists. Only tirzepatide itself was excluded. So this is not a drug against nothing; it is one drug against the realistic alternative, including its own closest competitors.

Non-inferiority was the primary objective and superiority followed on every secondary measure tested. Gastrointestinal effects were the commonest adverse events in both arms — unsurprising when the comparison arm could also contain a GLP-1.

Why 'normoglycemia' is not remission

The distinction has a practical edge rather than a semantic one. A person told their diabetes is reversed may reasonably conclude the drug has done its job and can stop. What the evidence on stopping shows is that blood sugar returns to roughly where it started within 12 to 18 months — covered in stopping as a clinical transition.

A normal number on treatment describes the treatment working. It does not describe a condition that has gone.

None of which diminishes the result. Two-thirds of people bringing an HbA1c into the non-diabetic range and holding it for two years, against a quarter on the realistic alternative, is a large effect on a meaningful outcome. It is simply an argument for continuing a treatment rather than for having finished one.

What we cannot tell from the published account

The strength of the comparison depends on what the conventional-care arm actually received, and the published abstract does not say. Guidelines permitted GLP-1 receptor agonists in that arm; whether most participants got one, or mostly got sulfonylureas and insulin, changes how impressive a 0.68-point gap is.

We are not going to guess in either direction. If the comparison arm was largely GLP-1-treated, this is a striking result against a strong bar. If it was largely older agents, the bar was lower than the design suggests. The full trial report will say; the abstract does not.

• Open-label, which the authors list as their own limitation. It is also unavoidable — you cannot blind an arm defined as “whatever the clinician chooses.”
• At most four years since diagnosis. This is early disease, where every agent performs best. Nothing here transfers to someone fifteen years in.
• Funded by Eli Lilly, who make tirzepatide. That is context, not disqualification, for a phase 4 trial in the Annals of Internal Medicine.
• Two years is long for this field and short for a lifelong condition. Durability past that point is not addressed.