The review searched peer-reviewed literature, clinical-trial registries, regulatory documents, and official sponsor sources for evidence available through August 2026. It covered approved, under-review, and investigational therapies, giving priority to randomized controlled trials, mechanistic studies, consensus and guideline documents, and human studies of biomarkers, treatment response, and omics-based stratification. Because this was a narrative review rather than a systematic review or meta-analysis, the authors did not register a formal protocol, conduct duplicate study selection or risk-of-bias assessment, or pool results quantitatively. Sponsor communications and registry records were used to describe development or regulatory status when information was not yet available in peer-reviewed publications, and such findings were treated as provisional.

Coexisting obesity and type 2 diabetes form a heterogeneous metabolic phenotype. Excess adiposity, particularly visceral and ectopic fat, can contribute to insulin resistance, inflammation, disturbed adipokine signaling, fatty liver, and impaired glucose regulation, while the two conditions together are associated with cardiovascular disease, chronic kidney disease, metabolic dysfunction-associated steatotic liver disease, heart failure, obstructive sleep apnea, osteoarthritis, and several cancers. People with the same body-mass index and hemoglobin A1c can nevertheless differ substantially in beta-cell reserve, liver fat, insulin resistance, appetite biology, inflammatory activity, skeletal muscle status, kidney function, cardiovascular vulnerability, and medication tolerance. These differences are relevant to treatment, but most have not been validated as predictive biomarkers for choosing one anti-obesity drug over another.

Clinical trials have established weight and glycemic benefits for several incretin-based therapies. Liraglutide demonstrated greater weight reduction than placebo in the SCALE Diabetes trial among adults with overweight or obesity and type 2 diabetes, and its lower antidiabetic dose reduced major adverse cardiovascular events in the LEADER trial. Tirzepatide, which activates GIP and GLP-1 receptors, has established efficacy for body weight and HbA1c. By August 2026, oral semaglutide 25 mg had received US Food and Drug Administration approval for chronic weight management, while the non-peptide oral GLP-1 receptor agonist orforglipron had also received approval. These results support the effectiveness of incretin therapies at population level; they do not show that a molecular test can identify which patient will derive the greatest benefit from one agent rather than another.

More complex drug candidates extend beyond single-receptor incretin therapy. Retatrutide is an investigational weekly GLP-1, GIP, and glucagon receptor agonist whose peer-reviewed phase 2 trial reported dose-dependent weight reduction approaching 24% at 48 weeks; sponsor-reported phase 3 results indicated a mean reduction of 28.3% at 80 weeks with the 12-mg dose. Survodutide, pemvidutide, and mazdutide combine GLP-1 activity with glucagon receptor activation, and amylin-based programs include CagriSema, amycretin, petrelintide, cagrilintide, and eloralintide. MariTide combines GLP-1 agonism with GIP antagonism and is being developed as a monthly injectable. These programs have different phases, populations, formulations, and levels of evidence, so their results should not be treated as interchangeable or ranked conclusively from indirect comparisons.

The biological rationale for these drugs extends beyond appetite suppression. Hormones including GLP-1, GIP, glucagon, peptide YY, and amylin connect the gut, brain, liver, adipose tissue, skeletal muscle, pancreas, kidney, and cardiovascular system. Genomics can capture inherited variation associated with insulin secretion, fat distribution, hepatic lipid metabolism, drug metabolism, and adverse-event susceptibility. Epigenomics may reflect the influence of obesity, hyperglycemia, aging, inflammation, and environmental exposures on metabolic tissues, while transcriptomics and proteomics can illuminate inflammation, tissue remodeling, and organ stress. Metabolomics and lipidomics can track small molecules and lipids associated with insulin resistance, lipotoxicity, and liver disease, and the microbiome may influence incretin secretion, bile-acid metabolism, intestinal barrier function, and gastrointestinal tolerability. These layers are complementary, but biological plausibility is not equivalent to validated clinical utility.

Human studies illustrate why stronger validation is required. In an analysis of 13,123 participants in the MetGen consortium, the SLC2A2 rs8192675 variant was associated with a 0.17-percentage-point greater HbA1c reduction with metformin; among participants with obesity, the CC versus TT genotype was associated with a 0.33-percentage-point greater reduction. A study of 233 adults with type 2 diabetes found that the TT genotype at GLP1R rs10305420 was associated with optimal glycemic response to liraglutide, with an odds ratio of 3.28, a 95% confidence interval of 1.41 to 7.65, and P=0.006. These are candidate pharmacogenetic markers rather than prescribing rules, because neither has been independently validated in a prospective trial showing that genotype-guided treatment improves outcomes. Cross-sectional metabolomic patterns associated with metformin or sulfonylurea response likewise remain exploratory.

A predictive biomarker must also be distinguished from a pharmacodynamic marker. A predictive marker identifies, before treatment, which therapy should provide greater benefit than an alternative. A pharmacodynamic marker instead records a biological change after treatment and may clarify how a drug acts. In 259 adults with type 2 diabetes participating in a phase 2b randomized trial, treatment-induced changes in branched-chain amino acid-related metabolites and lipid species correlated with improvements in HbA1c, HOMA2-IR, and proinsulin, but the study did not establish a pretreatment predictive effect. Multi-omics algorithms face further obstacles, including overfitting, batch effects, platform differences, reproducibility problems, poor calibration, and a lack of transportability across laboratories and populations. Artificial intelligence can integrate molecular, clinical, imaging, digital, and behavioral data, but model sophistication cannot substitute for prospective evidence that its recommendations improve care.

Several practical questions also remain unresolved. Weight loss is not synonymous with preserved muscle health: dual-energy X-ray absorptiometry-derived lean mass includes water and other non-fat tissues, not only contractile skeletal muscle. Trials therefore need more specific muscle assessments, including appendicular lean mass and muscle cross-sectional area or volume measured by computed tomography or magnetic resonance imaging, together with functional outcomes such as handgrip strength, chair-stand performance, gait speed, and the Short Physical Performance Battery. These measures are particularly important for older adults and people with frailty or sarcopenic obesity. Durability is another concern. Participants in the STEP 1 extension regained a substantial portion of previously lost weight after withdrawal of semaglutide 2.4 mg, while withdrawal from tirzepatide in SURMOUNT-4 was followed by weight regain and partial reversal of several cardiometabolic risk-marker improvements. Long-term evaluation must also address gastrointestinal adverse events, gallbladder disease, pancreatitis concerns, heart-rate changes, dehydration, malnutrition, and changes in muscle quantity or function.

The review argues that the next decisive step is not simply identifying more molecular associations, but testing whether they are predictive, reproducible, and clinically useful. Future trials should prespecify candidate biomarkers, formally test interactions between treatment and biomarker status, and include people with diverse ancestries and relevant conditions such as kidney disease, heart failure, fatty liver disease, frailty, and longer-standing diabetes. Metabolic endpoints such as weight, HbA1c, and time in range should be separated from intermediate markers such as liver fat, blood pressure, lipids, and albuminuria, and both should remain distinct from adjudicated cardiovascular and kidney outcomes. Combination, sequencing, induction-maintenance, and de-escalation strategies also require dedicated randomized trials. Until prospective studies show that molecularly guided prescribing improves efficacy, safety, persistence, patient-reported outcomes, or efficiency beyond high-quality conventional care, multi-omics remains a research framework for generating and testing precision-treatment hypotheses rather than a clinically validated component of routine prescribing.

Subject of Research: Multi-omics and anti-obesity drug selection in type 2 diabetes

Article Title: Diabesity: an old term, a modern challenge — multi-omics and precision anti-obesity pharmacotherapy in type 2 diabetes

Article References: Hoffmann, K., Pucek, W. F., Skrzypek, W., Kierzkowski, F., Ratajczak, P., Maggio, V., Rizzo, M., El-Tanani, M., Rabbani, S. A., Bhongade, B., Satyam, S. M., Talath, S., Rangraze, I. R., Wali, A. F., Ibraheem, W., El-Tanani, Y., Ispas, S., Bryl, W., & Paczkowska, A. (2026). Diabesity: an old term, a modern challenge — multi-omics and precision anti-obesity pharmacotherapy in type 2 diabetes. Journal of Pharmacy & Pharmaceutical Sciences, 29, Article 17560. https://doi.org/10.3389/jpps.2026.17560

Image Credits: AI Generated

DOI: 10.3389/jpps.2026.17560

Keywords: type 2 diabetes, obesity, anti-obesity pharmacotherapy, multi-omics, precision medicine, incretin therapies, biomarkers, clinical trials, body composition, Diabesity, term, modern

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