About 9% of your body weight, within a year. That is the within-person figure from a meta-analysis of 8,993 adults. A second meta-analysis, of 17 studies and 3,793 people in cessation arms, put the pooled regain at 7.20% [2]. Both are averages over wide spreads.
How much comes back depends on which drug came off. Pooled by molecule, liraglutide 3.0 mg gave back 4.83%, semaglutide 2.4 mg 7.19%, and tirzepatide 13.04% [2]. The ordering follows how much each took off. So the drug that works best costs the most to stop.
This paper publishes two percentages. They are not two versions of one answer. They measure different things, and only one of them answers what a buyer wants to know.
The number most people will quote
Six studies were pooled. In them, 5,553 adults stopped their GLP-1 and 3,440 stayed on it. [1] The difference in weight between those two groups came to 17.90%, with an interval of 14.11 to 21.69.
Read it carefully. It is a gap, and a gap has two sides. The people still taking the drug were still losing. The people who stopped were gaining. The 17.90% adds both together. It describes a distance between two futures, not a weight anybody put back on.
What happens when stopping is not a choice
Japan caps how long these drugs may be prescribed, which produces the one cohort where everybody stopped on schedule [3]. A prospective study weighed 104 patients through every phase of that rule. After the mandated stop, 95% had regained by two months. Indexed to the start, mean change moved from −14.4% at month 16 to −7.0% at month 22. Roughly half the achieved benefit was gone. The full reading of that cohort is in the article on Japan’s treatment cap.
Health economists have started pricing that pattern directly. A lifetime model of semaglutide 2.4 mg in Japan priced a quality-adjusted life year at ¥5,300,580, without type 2 diabetes [4]. It tested zero, one and two retreatment courses, because the cap forces a gap. Allowing retreatment improved the modeled gains substantially. Put the other way round, the economics assume you go back on it. That is a simulation, not a measurement.
What that means for a year of paying
Most of the loss is rented, not bought. That is the honest summary.
It also lands on an audience that mostly does not last a year. Retention data shows most people stopping inside six months. The regain figure applies to the majority, not an unlucky few.
The pace matters too, and this paper does not give one. For that you need the studies that tracked the curve month by month.
The six studies disagree
Heterogeneity tested at P = 0.0082. In plain terms, the six results are further apart than chance explains.
A pooled average across disagreeing studies is still useful. It is just less precise than the interval printed beside it suggests, and the interval is already fairly wide.
The authors also report larger rebound after tirzepatide than after semaglutide. No trial compared the two withdrawals head to head. That is a comparison assembled across separate studies, which is the weakest shape a drug comparison comes in.
What gets better when you stop
Side effects. Overall adverse reactions fell after withdrawal, and gastrointestinal ones fell specifically.
Cardiovascular event rates did not change either way. So stopping is not only a loss column, and a long prepay should not imply otherwise. Ask whether a year of this is worth its price and regain sits on one side, side effects on the other.
How to use it
Budget for the drug as an ongoing cost, not a course. Expect roughly 9% of your body weight back if you stop, on this evidence.
And if the plan is to stop at a target, plan the stop itself. Dose matters to whether people stay at all. That is its own decision. The dose people actually stay on is rarely the one on the label.