Why Outcomes Differ—and What Can Change Them

The average post-cessation GLP-1RA curve is a useful starting point, but it is not a personal forecast. The important question is why two people—or two studies—can show different amounts of regain, and what changes when withdrawal is supported, gradual, or forced by circumstances. Comparing those conditions helps readers judge maintenance claims and prepare informed questions for a clinician without turning population evidence into instructions for changing medication themselves.

The same treatment label can contain different trajectories

The amount lost before withdrawal is one obvious source of difference. In the STEP 1 extension, participants who lost more during the 68-week semaglutide period tended to regain more afterward, measured in absolute or percentage-point terms. Yet they also retained greater net loss at week 120 than participants who lost less. The pattern is internally consistent: a larger initial loss leaves more weight available to regain, while also leaving room for a larger net loss after that regain.

This finding should not be converted into a causal rule. The study’s subgroup comparisons were descriptive. They do not prove that losing more directly causes more later regain, nor do they predict the path for a particular person. Initial loss may be related to other factors, including treatment response, duration, support, or the characteristics of people who reach a given level of loss. The correct use of the finding is to expect that the size of the earlier loss matters when interpreting a later regain percentage.

Treatment duration and post-treatment support create another comparison problem. The supplied studies used different treatment periods and follow-up schedules, and some continued lifestyle counseling after withdrawal while others did not. Comorbidities and concomitant medications were also not reported consistently enough for the review to determine whether they altered the dynamics. A result after a long treatment period with continued support is not directly interchangeable with a result after a shorter period in which both medication and structured support ended.

The drug label itself is not a complete description of the intervention. The review includes liraglutide, semaglutide, tirzepatide, and other agents with different efficacy and pharmacokinetic properties. Newer semaglutide and tirzepatide studies generally produced greater on-treatment loss than older exenatide and liraglutide studies. Tirzepatide also acts on the GIP receptor as well as the GLP-1 receptor. Those differences matter because the amount and nature of the treatment response can change what happens at withdrawal.

At the same time, the selected modeled trajectories for liraglutide, semaglutide, and tirzepatide appeared broadly similar. That result argues against treating every drug as a completely separate phenomenon, but it does not establish identical post-cessation behavior. The review excluded a beinaglutide study from the trajectory model because its short-acting pharmacology was markedly different and less comparable. A class-level pattern can be useful for orientation while remaining too coarse for a drug-specific or person-specific prediction.