These drugs are discussed as though they were a lifetime therapy, and the clinical logic points that way. The randomized record does not stretch anywhere near a lifetime. Four trials carry the horizon out past a year, and they disagree with each other in instructive ways. The gap between what has been measured and what is being assumed is worth holding before costing a monthly price as a permanent line item.
Four years, in a cardiovascular population
The furthest-reaching weight analysis comes from SELECT, where treatment continued to 208 weeks. Weight loss continued over 65 weeks and was then sustained. At four years, semaglutide was associated with a mean reduction in weight of 10.2%, in waist circumference of 7.7 cm and in waist-to-height ratio of 6.9%, against 1.5%, 1.3 cm and 1.0% on placebo.[1]
That 10.2% is lower than the 68-week figure from the flagship obesity trial, and the reason is the population rather than any fading of effect. SELECT enrolled adults who already had cardiovascular disease, which is the subject of the outcome-trial article. Serious adverse events ran lower on semaglutide in every body-mass index category, while discontinuation of the trial product ran higher and rose as body-mass index class fell.[1]
Two years, in three hundred people
STEP 5 is the dedicated long-term obesity trial for semaglutide, and it is small. It randomized 304 adults, 152 to each arm, and 92.8% completed. Mean change in body weight from baseline to week 104 was −15.2% against −2.6% on placebo, a treatment difference of 12.6 percentage points (95% CI, −15.3 to −9.8), with 77.1% reaching a reduction of 5% or more against 34.4%.[2]
A two-year percentage resting on 152 treated participants is a reasonable estimate and a thin one. It is also the figure most often repeated as proof that results hold, so the sample size deserves to travel with it.
Tolerability at that horizon is worth reading too. Gastrointestinal adverse events were reported by 82.2% of the semaglutide group against 53.9% on placebo across the two years, mostly mild to moderate.[2]Those are cumulative figures over 104 weeks rather than a snapshot of any given week, which is a distinction most summaries drop.
Three years, and what the four months after showed
Tirzepatide reaches further. Among the 1,032 SURMOUNT-1 participants who entered with prediabetes as well as obesity, treatment ran 176 weeks followed by a 17-week off-treatment period. Mean change at week 176 was −12.3% at 5 mg, −18.7% at 10 mg and −19.7% at 15 mg, against −1.3% on placebo.[3]
The off-treatment window is the part worth reading twice. Over the 176 weeks, type 2 diabetes was diagnosed in 1.3% of treated participants against 13.3% on placebo, a hazard ratio of 0.07. After 17 weeks off treatment those figures were 2.4% and 13.7%, a hazard ratio of 0.12.[3] Protection did not vanish in four months, and it did narrow, which fits what the withdrawal trials found for weight itself.
The attrition sitting under every long number
The oldest three-year dataset in this class belongs to liraglutide, and it is candid about its own limits. SCALE randomized 2,254 adults with prediabetes and followed them to week 160. Of those, 1,128 completed, after 714 withdrawals in the liraglutide group and 412 in the placebo group.[4]
Half the trial had left before the finish line. Weight change at week 160 was −6.1% against −1.9%, and time to diabetes onset was 2.7 times longer on liraglutide, a hazard ratio of 0.21 (95% CI, 0.13 to 0.34).[4]The authors state plainly that withdrawn individuals were not followed up after discontinuation. Every long-horizon percentage in this field is computed on the people who stayed, and those people are, by definition, the ones for whom the drug was working and tolerable.
What is genuinely unknown
Four things, stated without hedging. Nothing among the trials on this page randomizes treatment past 208 weeks, so a claim about year six or year ten is an extrapolation rather than a result. None of them studied a compounded preparation, which is what most sellers on the review index actually dispense, so the durability evidence belongs to branded product at labeled doses.
None of them tested a taper, a drug holiday or an intermittent schedule against continuous treatment, so the strategies people improvise to lower their costs sit outside the evidence entirely — including the reduced weekly regimens described in the dose article. And because withdrawal from these trials was substantial, what happens to people who leave treatment unplanned is measured in far less detail than what happens to those who complete.
Reading a durability claim on a pricing page
A seller writing that results are “sustained” is usually pointing at one of the trials above, and the useful question is which one, for how long, and in whom. Two years in 304 people, four years in a cardiac population and three years in a prediabetes subgroup are three different promises wearing the same word.
A second question separates a durability claim from a retention claim. The trials report what happened to people who kept taking the drug, and a seller reporting the same thing about its own subscribers is describing who renewed rather than what the treatment did. Neither number is dishonest on its own. They answer different questions, and only one of them was randomized.
The commercial consequence is simple arithmetic. If the expectation is continuous treatment measured in years, the variable that compounds is the price per month at a maintenance dose, not the introductory offer. That is why a price that holds as the dose climbs outweighs a first-month discount. It is also why the compounded question gets larger the longer the horizon runs, since the durability evidence was earned by a product nobody in that market is shipping.