Background: Incretin-based anti-obesity pharmacotherapies deliver clinically meaningful weight loss, yet discontinuation is frequently followed by weight regain. Quantifying the velocity and uncertainty of post-withdrawal regain is operationally relevant for structured maintenance planning.

Methods: We conducted a secondary, cross-trial comparative analysis using published summary data from the STEP-1 trial extension (semaglutide 2.4 mg) and the SURMOUNT-4 randomized-withdrawal trial (tirzepatide; placebo arm after withdrawal). The primary endpoint was 12-month cumulative weight regain expressed as percent of withdrawal-baseline body weight. Secondary endpoints included regain expressed as percentage points vs original baseline, percent of lost weight regained, and monthly regain velocity. Uncertainty was quantified using confidence intervals derived from published variability measures and Monte Carlo propagation for derived ratios; between-therapy comparisons were exploratory.

Results: Over 12 months post-discontinuation, cumulative regain was approximately 14% of withdrawal-baseline body weight for both semaglutide and tirzepatide withdrawal (monthly velocity approximately 1.17% per month), with overlapping uncertainty bands. Expressed vs original baseline, regain was 11.6 percentage points after semaglutide withdrawal and approximately 11.1 percentage points (derived) after tirzepatide withdrawal. Semaglutide participants regained approximately two-thirds of weight lost during treatment, while tirzepatide withdrawal regained approximately one-half of lead-in loss. SURMOUNT-4 intermediate data suggested front-loaded regain, supporting interpretation of linear velocity as an average slope over 12 months.

Conclusions: When normalized to withdrawal-baseline weight, post-discontinuation regain velocity appears similar between semaglutide and tirzepatide withdrawal. These findings reinforce obesity as a chronic disease requiring structured maintenance and off-ramp strategies rather than abrupt discontinuation.

The obesity treatment paradigm has shifted materially with incretin-based pharmacotherapies. Semaglutide (GLP-1 receptor agonist) and tirzepatide (dual GIP/GLP-1 receptor agonist) have become cornerstone assets in chronic weight management programs, given their robust on-treatment efficacy and cardiometabolic benefits.A recurring execution risk, however, is weight recidivism after therapy discontinuation, attributable to biological and behavioral reversion once pharmacologic appetite and reward modulation is removed. From an operating-model perspective, clinicians require quantifiable and communicable regain trajectories (including uncertainty bounds) to design tapering, follow-up cadence, nutritional reinforcement, and behavioral maintenance pathways.This study provides a cross-trial, uncertainty-aware comparative analysis of 12-month weight-regain dynamics following discontinuation of semaglutide versus tirzepatide.

Data sources and study selection

Published summary outcomes were extracted from the STEP-1 trial extension (semaglutide 2.4 mg) and the SURMOUNT-4 randomized-withdrawal trial (tirzepatide). This analysis used aggregate data only; therefore, cross-trial comparisons are exploratory and intended for contextualization rather than head-to-head inference.

Endpoint definitions and denominator governance

To maximize comparability across trials, the primary endpoint was defined as 12-month cumulative weight regain expressed as percent of withdrawal-baseline body weight (Regain_WB). Withdrawal baseline corresponded to end-of-treatment weight at the point of discontinuation (week 68 for STEP-1 extension; week 36 randomization for SURMOUNT-4). Secondary endpoints included regain as percentage points vs original baseline (Regain_BL), percent of lost weight regained, and monthly regain velocity (linearized average slope across 12 months).

Linear projection model: rationale and assumptions

A linear projection was applied to translate 12-month regain into an average monthly velocity for operational use (patient counseling, visit cadence planning, and maintenance program design). The linear model was treated as a first-order approximation of cumulative regain over 12 months. Assumptions included: (i) the monthly slope represents an average over the interval (acknowledging potential non-linearity), (ii) no explicit modeling of time-varying regain drivers or structured interventions, and (iii) cross-trial comparability is limited by design and population differences.

Statistical analysis and uncertainty quantification

Where SD and sample size were reported (STEP-1 extension), 95% confidence intervals for means were calculated using normal approximation (mean +/- 1.96 x SD/sqrt(n)). Where least-squares means and 95% confidence intervals were reported (SURMOUNT-4), standard errors were reconstructed from the confidence interval width. For derived ratios (e.g., Regain_WB for STEP-1), uncertainty was propagated using Monte Carlo simulation. An exploratory comparison of monthly velocities between therapies was performed using a z-test on the difference in means; results were interpreted as hypothesis-generating.

Extracted and derived metrics

STEP-1 extension (semaglutide): on-treatment loss 17.3% (SD 9.3); post-withdrawal regains 11.6 percentage points (SD 7.7) at 52 weeks; net change vs baseline -5.6% (SD 8.9). SURMOUNT-4 (withdrawal/placebo arm): lead-in loss 20.9%; change from week 36 to week 88 +14.0% (95% CI 12.8-15.2); intermediate regain from week 36 to week 64 +10.0% (95% CI 9.0-11.0).

Primary endpoint: regain normalized to withdrawal-baseline body weight

At 12 months post-discontinuation, Regain_WB was approximately 14% of withdrawal-baseline body weight for both semaglutide and tirzepatide withdrawal, corresponding to an average monthly regain velocity of approximately 1.17% per month. Uncertainty intervals were overlapping, and exploratory testing did not indicate a meaningful difference in average monthly velocity [1-10].

Secondary endpoints

Expressed as percentage points vs original baseline, regain was 11.6 percentage points after semaglutide withdrawal and approximately 11.1 percentage points after tirzepatide withdrawal (derived). Semaglutide participants regained approximately two-thirds of weight lost on-treatment, while tirzepatide withdrawal regained approximately one-half of lead-in loss.

Linearity check

SURMOUNT-4 reported an intermediate withdrawal timepoint showing 10.0% regain by approximately 6.5 months and 14.0% regain by 12 months, suggesting front-loaded regain. Accordingly, the linear model should be interpreted as an average slope over 12 months rather than a time-invariant mechanistic constant.

This cross-trial, aggregate-data analysis was designed to translate post-discontinuation outcomes from STEP-1 extension and SURMOUNT-4 into operationally interpretable regain metrics, including a normalized monthly velocity and uncertainty-aware interpretation. The primary objective is not to establish head-to-head superiority, but to provide a structured framework for counseling, follow-up cadence design, and maintenance/off-ramp planning after discontinuation of incretin-based anti-obesity therapy.

Interpretation of the primary findings

Across both programs, discontinuation was followed by clinically meaningful regain over 12 months. When regain is expressed as percent of body weight at the time of withdrawal (withdrawal-baseline normalization), the magnitude of cumulative regain over one year is broadly comparable between semaglutide withdrawal and tirzepatide withdrawal. This denominator choice is strategically important because it anchors the analysis to the discontinuation event and avoids conflating post-withdrawal regain with differences in initial treatment efficacy. By contrast, when regain is expressed in absolute terms (percentage points vs original baseline or kilograms), tirzepatide can appear to "rebound more" largely because it produces greater on-treatment weight loss, increasing the absolute amount of weight at risk once pharmacologic appetite modulation is removed. From an operating-model perspective, this implies that patients achieving larger treatment responses may require more robust discontinuation governance - structured taper, closer monitoring, and intensified behavioral reinforcement - primarily because the absolute rebound envelope is larger, not necessarily because the underlying withdrawal biology is fundamentally different.

What the linear model does - and does not – claim

A central methodological requirement is to clarify that the linear projection used here functions as a first-order approximation. It converts a 12-month cumulative regain estimate into a monthly velocity metric that can be communicated in routine practice and used for pathway design. The linear model should therefore be interpreted as an average slope across the follow-up interval, not as a mechanistic claim that regain occurs at a constant rate. Where intermediate timepoints are available (notably in SURMOUNT-4), the data suggest that regain may be front-loaded - higher early regain followed by relative deceleration later - consistent with a non-linear trajectory. This supports an operational inference that the first 3-6 months after discontinuation represent a high-leverage window for relapse prevention. Accordingly, the linear model is best positioned as a planning tool for cross-program comparability and patient counseling, while acknowledging that time-varying regain is likely and may be better captured by piecewise or saturating models when sufficient timepoints exist.

Keeping claims within the data: uncertainty and inferential framing

In cross-trial aggregate analyses, it is essential to separate the magnitude of the observed phenomenon from small between-treatment differences that may be confounded by design and population heterogeneity. When uncertainty is considered, the practical conclusion remains that withdrawal-associated regain is substantial across both agents. Any inferred differences in normalized regain velocity, if present, should be treated as hypothesis-generating rather than definitive. This framing directly informs the clinical message: the priority is not marginal distinctions in rebound kinetics, but rather the implementation of durable maintenance architectures that treat discontinuation as a managed transition. In this context, regain velocity is a risk-stratification and communication tool, not a proof of mechanistic divergence.

Mechanistic considerations: literature-consistent hypotheses, not causal outputs

Mechanistic explanations for regain are clinically relevant but must be positioned appropriately. The current analysis does not measure appetite hormones, energy expenditure, satiety signals, or behavioral adherence; therefore, mechanistic statements cannot be presented as findings derived from the model. Instead, they should be framed as plausible, literature-consistent hypotheses that provide context for why regain is expected after withdrawal. Within this framing, post-withdrawal regain plausibly reflects the restoration of appetite drive and food reward signaling, reversal of delayed gastric emptying and satiety effects, and counter-regulatory neuroendocrine responses that promote energy repletion. These phenomena are consistent with the broader physiology of body-weight homeostasis and the chronic-relapsing nature of obesity. However, they remain contextual explanations rather than causal outputs of this comparative modeling exercise.

Clinical and operational implications: from "stop" to "off-ramp"

A key clinical implication of these findings is that discontinuation should be managed as an off-ramp rather than an endpoint. The concept of a monthly regain velocity enables clinicians to define early-warning thresholds and establish a structured monitoring plan. This is particularly relevant given the potential for early accelerated regain suggested by intermediate withdrawal timepoints. Operationally, a pragmatic discontinuation pathway should emphasize early intensification of support, including increased follow-up cadence in the first months, structured nutritional planning favoring high-satiety patterns, behavioral relapse-prevention strategies (self-monitoring, stimulus control, goal reinforcement), and - where clinically appropriate - consideration of maintenance dosing, tapering, or transition strategies rather than abrupt cessation. These actions are framed as risk-mitigation steps justified by the magnitude and timing of regain, not as interventions validated by causal inference within this dataset.

Limitations and next steps

This analysis has important limitations. First, the absence of head-to-head withdrawal trials means that cross-trial comparisons are vulnerable to confounding by differences in populations, trial conduct, and background lifestyle programs. Second, the reliance on aggregate data limits the ability to model inter-individual variability, subgroup effects, and time-varying regain dynamics with precision. Third, the linear model simplifies regain into an average slope and may understate early rapid regain; future work should incorporate piecewise or asymptotic modeling when more timepoints are available. Next-step evidence priorities include real-world longitudinal datasets with standardized discontinuation definitions, matched withdrawal cohorts, and modeling frameworks that jointly estimate weight change and mediators (e.g., appetite, dietary intake proxies, resting energy expenditure) to strengthen causal interpretation and enable better individualized off-ramp planning (Figure 1 and Table 1).

Figure 1: Post-discontinuation weight-regain dynamics: observed points and 12-month linear projections.

Table 1: Cross-trial summary and derived regain metrics (12 months post-discontinuation).

In summary, discontinuation of semaglutide and tirzepatide is followed by substantial weight regain within 12 months. When normalized to withdrawal-baseline body weight, regain appears broadly similar across agents, reinforcing the interpretation of rebound as a withdrawal phenomenon consistent with chronic weight homeostasis. The most actionable conclusion is the need for structured maintenance and off-ramp pathways to preserve long-term outcomes.

When denominators are standardized to withdrawal-baseline body weight, semaglutide and tirzepatide withdrawal demonstrate similar 12-month regain magnitude and average monthly velocity, with evidence of front-loaded regain in SURMOUNT-4. Discontinuation should be operationalized as a structured transition via maintenance programs, taper/off-ramp protocols, and sustained lifestyle support to preserve durable outcomes. 

1. Wilding JPH, Batterham R, Calanna S, Davies M, Gaal LFV, Lingvay I, et al. Once-Weekly Semaglutide in Adults with Overweight or Obesity. N Engl J Med. 2021; 384: 989-1002.
2. Wilding JPH, batterham RL, Davies M, Van Gall LF, Kandler K, Konakli K, et al. Weight regains and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension. Diabetes Obes Metab. 2022; 24: 1553-1564.
3. Jastreboff AM, Aronne LJ, Ahmad NN, Wharton S, connery L, Alves B, et al. Tirzepatide Once Weekly for the Treatment of Obesity. N Engl J Med. 2022; 387: 205-216.
4. Aronne LJ, Sattar N, Hom DB, Bays HE, Wharton S, et al. Continued Treatment with Tirzepatide for Maintenance of Weight Reduction in Adults with Obesity: The SURMOUNT-4 Randomized Clinical Trial. JAMA. 2024; 331: 38-48.
5. Kushner RF. Effects of Withdrawal of Obesity Pharmacotherapy on Weight Regain: A Systematic Review and Meta-Analysis. Obesity Reviews. 2023; 24: e13502.
6. American Diabetes Association. Pharmacologic Approaches to Glycemic Treatment - Standards of Care in Diabetes 2024. Diabetes Care. 2024; 4: S125-S143.
7. Garvey WT. Management of Obesity with Anti-Obesity Medications: A Clinical Framework. Nat Rev Endocrinol. 2024; 20: 415-432.
8. Wadden T. Weight Regain after Discontinuation of Obesity Medications: A Critical Review. Int J Obes. 2024; 48: 1101-1112.
9. Hall KD, Guo J. Obesity Energetics: Body Weight Regulation and Metabolism. Nat Rev Endocrinol. 2021;17: 323-332.
10. Bray GA, Ryan DH. Evidence-Based Approach to Obesity Pharmacotherapy. J Clin Endocrinol Metab. 2023; 108: 1237-1250.