The study, led by Chia Siang Kow and colleagues, took a fresh statistical look at the six clinical studies and ten intervention arms that tracked semaglutide or tirzepatide after treatment discontinuation, drawing on data covering 1776 participants. Rather than relying on simple pooled averages, the researchers reconstructed the arm-level data and applied a Bayesian hierarchical longitudinal model, a statistical framework that jointly models repeated measurements over time while explicitly accounting for variability between study arms. This approach allowed the team to go beyond asking how many kilograms are regained each month and instead translate the trajectory into clinically meaningful milestones, complete with full probability distributions that quantify uncertainty.

The headline finding is that people who stop taking these medications regain an estimated 1.04 kilograms per month on average, with a 95 percent credible interval of 0.80 to 1.29 kilograms per month. The modelled average participant had lost 15.35 kilograms by the time treatment ended. Under the linear model assumed by the researchers, half of that hard-won loss was projected to return within about 7.5 months, and participants were projected to be back at their baseline weight by roughly 15 months after stopping. In practical terms, the clock on the treatment’s benefits starts ticking almost the moment the final injection wears off.

The month-by-month trajectory is particularly striking. Within the follow-up window actually observed in the trials, which extended to 52 weeks, the modelled average weight change was still below baseline at six months, at minus 9.12 kilograms, but by then approximately 41 percent of the initial weight loss had already been regained. At nine months the average was minus 6.01 kilograms, corresponding to 62 percent regained, and by twelve months the average stood at minus 2.90 kilograms, meaning roughly 83 percent of the lost weight had returned. The posterior probability that the average trajectory had regained at least half of the initial weight loss climbed from just 12.5 percent at six months to 86.7 percent at nine months and a near-certain 99.5 percent at one year.

It is important to understand the mathematical machinery behind these numbers. The researchers assumed a constant linear regain slope rather than a curving trajectory, a choice justified because follow-up data were sparse and a prior systematic review had found that adding a nonlinear term did not improve model fit. Each study arm was treated as a repeated-measures trajectory with its own random effects for both the weight loss present at discontinuation and the subsequent regain rate. The Bayesian estimation relied on full Markov chain Monte Carlo sampling, with convergence confirmed by R-hat statistics hovering at approximately 1.00 and large effective sample sizes for the key slope parameters. Estimates extending beyond the maximum observed follow-up of 52 weeks, including the projected return to baseline at 15 months, are explicitly flagged as extrapolations of the average trajectory rather than directly observed outcomes.

One of the most provocative aspects of the analysis is its comparison of the two drugs. Tirzepatide arms showed a numerically faster unadjusted regain rate of 1.10 kilograms per month compared with 0.89 kilograms per month for semaglutide, and projected return to baseline was correspondingly sooner, at roughly 14.5 months versus 17.3 months. But when the researchers adjusted for the magnitude of initial weight loss and post-discontinuation support in a Bayesian meta-regression, the drug difference essentially vanished. The adjusted effect of tirzepatide versus semaglutide was a negligible minus 0.04 kilograms per month, with a posterior probability of only 40.2 percent that tirzepatide regains faster. In other words, the apparent difference in rebound between the two drugs likely reflects differences in how much weight was lost in the first place, not any inherent difference in the physiology of regain.

That observation points to one of the study’s most interesting exploratory findings: greater initial weight loss was itself the strongest directional predictor of faster absolute regain. Each additional 5 kilograms of weight lost during treatment was associated with a 0.20 kilograms per month faster regain slope, a result carrying a 92.1 percent posterior probability, though the credible interval included zero. The authors caution that this association may partly reflect mathematical coupling, since a larger initial loss simply creates more opportunity for absolute regain. Meanwhile, behavioural or lifestyle support after discontinuation showed a directional association with slower regain, an estimated effect of minus 0.19 kilograms per month with an 87.2 percent probability of benefit, but the imprecise estimate means the finding remains inconclusive. With only ten intervention arms available, all meta-regression results are explicitly exploratory and hypothesis-generating rather than definitive.

Sensitivity analyses reinforced the robustness of the core conclusion. A contrast analysis using the randomised differences between intervention and control arms showed a similar direction of effect, though with less precision. A post hoc analysis excluded the three SURPASS-1 arms, which came from a trial of adults with Type 2 diabetes receiving tirzepatide as glucose-lowering monotherapy rather than for weight management. Excluding those arms left the monthly regain slope essentially unchanged at 0.95 kilograms per month, though the projected time to 50 percent regain lengthened to 9.42 months because the average initial weight loss among the remaining arms was greater. The qualitative message of rapid regain survived every stress test the researchers applied.

The clinical implications are considerable. Because discontinuation is common in routine practice, driven by cost, tolerability problems, access restrictions, treatment fatigue and genuine uncertainty about how long therapy should last, the findings suggest that clinicians should plan proactive monitoring within the first several months after stopping treatment, when early regain can be identified and maintenance strategies reassessed. The authors are careful to note that their data do not determine whether dose tapering, lower-dose maintenance, drug switching, intermittent treatment or any specific behavioural programme can prevent regain; those questions require dedicated trials. They also stress that the analysis captured body-weight trajectories only, so any statements about cardiometabolic benefits, cost-effectiveness or the consequences of fixed-duration treatment policies remain hypotheses informed by the weight pattern rather than direct findings.

What the study ultimately delivers is a statistically rigorous quantification of something patients and clinicians have long suspected: incretin-based anti-obesity medications suppress appetite and enable weight loss while they are active, and withdrawing that physiological support removes the very mechanism holding the weight down. Between-study heterogeneity was real but moderate, with a between-arm standard deviation of 6.78 kilograms for initial weight loss and 0.26 kilograms per month for the regain slope, suggesting that while individual experiences vary, the average trajectory is consistently steep. As these drugs reshape obesity medicine and public expectations, the analysis argues that their long-term value must be judged not only by the dramatic losses achieved during treatment but by an honest accounting of what follows when the treatment ends, and by research that extends follow-up beyond one year, reports regain in both absolute and percentage terms, and directly tests the maintenance strategies that patients will increasingly demand.

Subject of Research: Weight regain after discontinuation of the incretin-based anti-obesity medications semaglutide and tirzepatide

Article Title: Weight Regain Trajectories After Discontinuation of Semaglutide or Tirzepatide: A Reconstructed Aggregate‐Data Bayesian Longitudinal Meta‐Analysis

Article References: Kow, C. S., Thiruchelvam, K., Ramachandram, D. S., & Zaihan, A. F. (2026). Weight Regain Trajectories After Discontinuation of Semaglutide or Tirzepatide: A Reconstructed Aggregate‐Data Bayesian Longitudinal Meta‐Analysis. Endocrinology, Diabetes & Metabolism, 9(5), Article e70325. https://doi.org/10.1002/edm2.70325

Image Credits: AI Generated

DOI: 10.1002/edm2.70325

Keywords: semaglutide, tirzepatide, weight regain, obesity, incretin therapies, GLP-1 receptor agonists, Bayesian meta-analysis, treatment discontinuation, weight loss maintenance, anti-obesity pharmacotherapy, Weight, Regain