Both arms of this decision are reconstructed from outcome and attitude data, not from anyone asking patients whether they regret the choice they made. There is no published regret survey for GLP-1 continuation as of mid-2026, so the numbers here are proxies: weight regain stands in for regret over stopping, and the reported difficulty of paying for the drug stands in for the burden of staying. Read them as the rate at which each path produces the thing people say they wanted to avoid, not as a count of stated regret.
Stopping at goal weight reverses most of the loss for most people. In the STEP-1 trial extension, participants who came off semaglutide regained a mean of two-thirds of their prior weight loss over the following year, and a 2026 meta-regression of six randomized trials put the figure at 60% regained by 52 weeks, plateauing near 75% of the original loss. Real-world data is gentler but points the same way: in a Cleveland Clinic cohort of 7,938 adults with obesity who stopped semaglutide or tirzepatide, 55% gained weight in the year afterward while 45% kept losing or held steady. The trial figures describe how much weight comes back; the 55% is the share of individual stoppers who saw the scale move the wrong way, which is why it, rather than the regain magnitude, anchors the action side.
Staying on carries a different cost, and a large share of long-term users describe it as hard to bear. In KFF’s late-2025 polling, 56% of GLP-1 users said the drug was difficult to afford, a share barely lower among the insured. The burden is concrete enough to end treatment for most people who start: a clinical-practice study found 54.9% discontinued within the first year, with cost or insurance driving 47.6% of those exits and side-effect intolerance another 14.6%. Those discontinuation figures are not folded into the regret rate here, because someone who quits has effectively taken the stopping path rather than the staying one. They establish only that the open-ended expense and tolerability of indefinite use is severe enough to push a majority off within twelve months.
The two proxy rates land almost on top of each other, 55% for stopping and 56% for staying, so the comparison is close to balanced rather than favoring either path. The symmetry is partly an artifact of proxy choice: a magnitude-of-regain framing would tilt the stopping side higher, and a stated-regret survey, if one existed, could move both. What the data does support is narrower and firmer. Stopping usually means regaining most of the loss, and staying usually means an expense a majority find difficult to sustain. Neither path offers a clean exit from the trade-off.