The strongest number says yes and the second-strongest says wait. Across 176 weeks of SURMOUNT-1, type 2 diabetes was diagnosed in 1.3% of the tirzepatide group against 13.3% on placebo, a hazard ratio of 0.07 [2]. Against that sits the screening arithmetic: of the people a one-hour glucose test labels high risk, roughly 85 in every 100 never develop diabetes at all over three years [1]. Both figures come from the same trial, and whether the drug is worth buying on this indication depends on which of them applies to the reader, alongside what prevention costs to buy.
What high risk actually predicts
A post hoc analysis tested three ways of identifying adults with obesity and prediabetes who were about to develop diabetes, scored against what happened in the placebo group over three years. A one-hour glucose reading of 155 mg/dL or above caught 83% of the people who went on to develop diabetes, and flagged so many others that only 15% of those it identified actually did. Fasting glucose failed in the opposite direction, missing half the cases while being right about 37% of the people it flagged. HbA1c sat between them at 19%.
The property nobody quotes is the useful one. All three tests had negative predictive values between 0.90 and 0.92, so a reassuring result is fairly reliable. These are instruments for ruling out rather than ruling in, and that asymmetry matters before a screening number is treated as a reason to start paying for something. Who reaches these drugs at all is counted in who the price filtered out.
What the drug prevented, and for how long
SURMOUNT-1 randomized 2,539 people with obesity, of whom 1,032 also had prediabetes, and followed that subgroup for 176 weeks of treatment plus 17 weeks off it [2]. At 176 weeks the mean weight change was −12.3% at 5 mg, −18.7% at 10 mg and −19.7% at 15 mg, against −1.3% on placebo. Diabetes was diagnosed in 1.3% against 13.3%, with a 95% CI on the hazard ratio of 0.0 to 0.1.
Seventeen weeks without the drug changed the picture. In that period the figures moved to 2.4% and 13.7%, a hazard ratio of 0.12 rather than 0.07, and the gap had begun closing immediately. A prevention effect measured on treatment is a recurring purchase rather than a course, which is the same structure described in what comes back after stopping.
The dose question, which is really the price question
SURMOUNT-MAINTAIN took 378 adults who had already spent 60 weeks on tirzepatide and randomized them to continue at their maximum tolerated dose, drop to 5 mg, or switch to placebo [3]. At week 112 weight sat 21.9% below baseline on the full dose, 16.6% at 5 mg and 9.9% on placebo.
Rescue therapy is the figure that translates. It was needed by 8% of the full-dose group, 25% of those who halved, and 67% of those switched to placebo. For someone buying to prevent diabetes rather than to lose a set number of pounds, the arithmetic changes. A cheaper maintenance dose that holds most of the benefit is a more realistic plan than paying full price indefinitely or stopping, as the maintenance options set out.
Reading the two sets of numbers together
One technical point separates figures that look comparable. The predictive values were calculated in the placebo group and describe natural history, while the prevention percentages come from the treated arms. A seller quoting both in a single sentence would be blending what happens anyway with what the drug does.
The practical reading is this. In a group where 15% would have developed diabetes without treatment, a 90% relative reduction prevents far fewer cases than the percentage suggests, and the treatment has to continue for the reduction to hold. Whether that is worth a recurring bill is a budget question rather than a clinical one, and what a GLP-1 costs per month is the other half of it.