Obesity Treatment is a Systems Failure: Why GLP-1 Drugs Will Plateau Without Personalization…

Obesity Treatment is a Systems Failure: Why GLP-1 Drugs Will Plateau Without Personalization Infrastructure

This essay has been submitted to Astera. If you are stuck behind a paywall then please click this friend link to read the article.

A 42-year-old woman begins semaglutide therapy. Within three months, she loses 8 kilograms — an outcome that would have seemed improbable just a decade ago. By month six, she’s plateaued. Nausea persists. Doses are missed. By month nine, she stops treatment altogether, and within a year, most of the weight has returned.

This is not an exceptional case. In real-world practice, it’s closer to the norm.

GLP-1 receptor agonists have been widely — and justifiably — described as transformative drugs for obesity. In controlled trials, they produce sustained and clinically meaningful weight loss. But real-world data tells a starkly different story. In a large cohort study of over 125,000 patients initiating GLP-1 receptor agonists, 64.8% of those without type 2 diabetes discontinued therapy within one year. A separate real-world analysis found that only 27.2% of commercially insured patients remained adherent through the first year, with two-thirds classified as non-persistent. The gap between what these drugs can do in a trial and what they actually deliver in clinic is not a pharmacology problem. It is a systems problem.

The Bottleneck: Treating a Systems Disease with a Single Intervention

Obesity is a complex, multi-layered condition shaped by genetic predisposition, neuroendocrine regulation, gut microbiome composition, and behavioral dynamics. GLP-1 therapies target a subset of these pathways — primarily appetite regulation and insulin signaling. They are, by design, partial interventions.

Yet in practice, they are deployed as if they were complete solutions.

The core problem is this: we’ve invested heavily in making the molecule precise while leaving the delivery system entirely generic. The healthcare infrastructure surrounding GLP-1 therapies lacks the capacity to personalize, adapt, and sustain treatment over time. Even a highly effective drug underperforms when the system around it is poorly designed.

The consequences show up across multiple dimensions simultaneously. Side effects are the single largest documented reason for GLP-1 discontinuation, accounting for 28.2% of all cessation events in real-world data. Many patients never reach optimal therapeutic doses because titration is managed reactively rather than proactively. Response is highly variable — some patients achieve substantial weight loss while others see minimal benefit on identical regimens. And perhaps most tellingly, the STEP 1 trial extension demonstrated meaningful weight regain in the year following semaglutide cessation, with participants recovering a significant proportion of what had been lost during 68 weeks of treatment. The drug works while you’re on it — but it doesn’t change the underlying trajectory. All of these failure modes trace back to the same gap: no mechanism exists to account for the biological and behavioral variability between patients.

The Response Heterogeneity Problem

One of the most consistent observations in clinical practice is how differently patients respond to GLP-1 therapy — not just in how much weight they lose, but in how well they tolerate the drug. Some patients achieve substantial, sustained weight loss with minimal side effects. Others plateau early, struggle with nausea at even modest doses, and discontinue within months. This variability isn’t random — it almost certainly reflects underlying differences in genetics, gut physiology, and metabolic state.

Despite this, GLP-1 therapies are prescribed using standardized dosing and titration protocols. There is no routine stratification based on biological markers, and no clinically actionable pharmacogenomic framework to guide therapy selection or dose escalation.

This is striking when you compare it to how oncology and psychiatry have evolved. In those fields, pharmacogenomics is now routinely used to inform drug selection and expected tolerability. Obesity medicine, by contrast, remains largely empirical. Clinicians initiate therapy, escalate doses, and manage adverse effects through trial and error — which is a poor fit for a chronic, heterogeneous condition in a healthcare system that rewards throughput over continuity.

The biological relevance is clear. What’s missing is the data infrastructure to act on it. Gut microbiota influence gastrointestinal mucosal permeability, bile acid metabolism, short-chain fatty acid synthesis, and incretin signaling — all pathways directly relevant to GLP-1 pharmacodynamics. Emerging evidence suggests the gut microbiome may be one of the more tractable handles we have on predicting response. A 2024 preprint from Klemets and colleagues showed that the baseline fecal microbiome predicted glycemic response to semaglutide, suggesting pre-treatment microbial profiling could have real clinical utility. A 2025 systematic review of 38 studies confirmed that GLP-1 analogues have a notable impact on the composition, richness, and diversity of the gut microbiota — though the directionality and clinical implications remain inconsistent across study populations, underscoring how much we still have to learn about this interaction.

There are currently no large-scale, longitudinal datasets linking genomics, microbiome profiles, dosing patterns, and clinical outcomes in GLP-1-treated populations. Without that kind of integrated data, pharmacogenomic insights cannot be translated into clinical decisions. We’ve built a precise drug and left it operating inside a blunt system.

Fragmented Diagnostics and the Missing Decision Layer

Even where relevant data exists, it’s scattered. Pharmacogenomic testing, microbiome analysis, and standard laboratory measurements are typically conducted in isolation, at different time points, by different providers, and interpreted without any shared framework. Their outputs rarely converge into a unified picture.

This fragmentation creates a critical gap: the absence of a clinical intelligence layer capable of synthesizing multi-dimensional data into something actionable. Without it, clinicians are left to either interpret disconnected data streams or simply ignore them. In GLP-1 therapy, this means genetic predispositions are not used to anticipate response or adverse effects; microbiome profiles are not leveraged to optimize metabolic outcomes; and longitudinal patient data is not fed back into treatment decisions. There’s no mechanism to carry forward what was learned from one patient encounter to the next. The system treats every visit as a clean slate.

This is not a feature — it’s a structural failure that becomes especially costly in a disease that requires years of adaptive management.

Episodic Care in a Chronic Adaptive Disease

A further structural limitation lies in how care itself is organized. Obesity is a chronic, adaptive condition that requires continuous management — not just prescription renewal. But healthcare systems are built around episodic interactions: clinic visits, lab orders, prescription fills. This rhythm is poorly matched to therapies that need ongoing titration, proactive side-effect monitoring, and sustained behavioral reinforcement.

Data from an academic multidisciplinary obesity clinic — one of the more structured real-world settings — found median GLP-1 persistence of only 10.7 months even with consistent insurance coverage and dietitian support. Yet this was markedly better than average, and the authors attribute the improvement precisely to the structured support environment. Without that kind of scaffolding, adherence collapses faster. The early success pattern is predictable: patients respond well initially, disengage as side effects accumulate or enthusiasm fades, and discontinue before reaching meaningful or durable outcomes.

A large retrospective cohort using Cleveland Clinic data found that patients who discontinued semaglutide or tirzepatide within three months achieved only 3.6% weight reduction at one year, compared to 11.9% in those who remained on therapy. The dose and duration needed to achieve trial-level outcomes are rarely reached in practice — not because the drug can’t get there, but because the surrounding care structure isn’t built to sustain the journey.

Reframing the Problem: From Drug to Infrastructure

Across all of these dimensions, the picture is consistent. The limiting factor in GLP-1 therapy is not drug efficacy — it’s system design. The current model lacks three things that are all, individually, necessary and none of which is sufficient on its own:

First, personalization infrastructure to account for the biological variability between patients — including genomic, microbiome, and metabolic inputs that currently go unmeasured and unused. Second, integration layers to synthesize multi-omic and clinical data into a coherent decision-making framework, rather than leaving clinicians to interpolate between siloed reports. Third, continuous care systems designed to sustain engagement, adapt treatment in real time, and prevent the kind of early discontinuation that renders the therapy ineffective.

That shift in framing matters beyond obesity. Many areas of medicine face the same structural problem: highly effective interventions are undermined by biological variability, fragmented data, and care architectures that can’t sustain the complexity of chronic disease management.

A Testable Solution: The PeptideRx Model

To address this gap, I propose a system-level intervention — a multi-layer personalization platform for metabolic disease management, which I’m calling the PeptideRx model. Rather than treating GLP-1 therapy as a standalone prescription, this model embeds it within an integrated three-layer framework.

Layer 1 — Biological Personalization. Before therapy initiation, patients undergo structured baseline profiling: pharmacogenomic panel, gut microbiome sequencing (16S rRNA or shallow shotgun), and metabolic phenotyping. These inputs are used to stratify predicted response and tolerability, identify risk factors for adverse effects, and inform starting dose and titration strategy — moving away from one-size-fits-all escalation toward a biologically anchored starting point.

Layer 2 — Clinical Optimization. Treatment is adjusted dynamically based on longitudinal data: weight trajectories, side-effect burden, metabolic markers, and adherence patterns. An algorithmic decision-support layer assists clinicians in managing dose escalation, flagging early signs of non-response, and recommending proactive rather than reactive adjustments. This layer essentially gives the care system memory — something the current model conspicuously lacks.

Layer 3 — Behavioral Engagement. Patients are supported through continuous, lightweight monitoring and feedback: appetite and dietary pattern tracking, adherence nudges calibrated to individual behavior patterns, and personalized lifestyle recommendations. The goal here is not to add burden, but to replace the passive gaps between clinic visits with something that sustains momentum.

Together, these layers transform GLP-1 therapy from a static prescription into a dynamic, adaptive clinical system.

Experimental Design

A straightforward pragmatic trial could test this. Two cohorts of patients initiating GLP-1 therapy would be enrolled and followed for 12 months: a control group receiving standard care, and an intervention group receiving PeptideRx-integrated care with baseline profiling, decision support, and behavioral engagement.

Primary outcomes would include percentage weight loss at six and twelve months, persistence and adherence rates, incidence and severity of adverse effects, and cost per successful outcome. Secondary analyses would explore correlations between baseline genomic or microbiome features and treatment response — contributing incrementally to the pharmacogenomic evidence base.

Even if early genomic signals are weak or inconsistent in the first trial cohort, the system generates the longitudinal, multi-omic dataset needed to refine personalization strategies over time. The intervention is therefore not only therapeutic but generative — it accelerates learning from real-world data in a way that isolated clinic visits never will.

Why This Matters Beyond Obesity

While this analysis focuses on GLP-1 therapies, the underlying problem is not unique to obesity. Psychiatry has wrestled with the same antidepressant response heterogeneity for decades. Oncology built precision medicine infrastructure precisely because empirical prescribing was leaving too much on the table. In both cases, the field improved not just by discovering better molecules, but by building the infrastructure to know which molecules to use, in whom, and for how long.

Obesity medicine is at an analogous inflection point. We have the molecules. What we lack is the surrounding system that enables them to work reliably across diverse, real-world populations — and that generates the knowledge we need to keep improving.

Conclusion

GLP-1 receptor agonists have genuinely shifted what’s possible in obesity treatment. But their real-world impact is being systematically constrained by structural limitations that have nothing to do with the pharmacology. High discontinuation rates, variable responses, and poor durability are not failures of the drugs themselves — they are failures of the systems in which those drugs are deployed.

Addressing this requires a fundamental shift in where we direct our attention. The harder problem isn’t finding better drugs. It’s building the infrastructure to make existing drugs actually work across diverse, real-world patients — patients who don’t look like trial participants, who stop filling prescriptions, and who need something more than a prescription to stay the course.

The next frontier in obesity medicine is not molecular. It’s architectural.

References

1. Rodriguez PJ, et al. Discontinuation and Reinitiation of Dual-Labeled GLP-1 Receptor Agonists Among US Adults With Overweight or Obesity. JAMA Network Open. 2025;8(1). doi:10.1001/jamanetworkopen.2024.xx
2. Gleason PP, et al. Real-world persistence and adherence to glucagon-like peptide-1 receptor agonists among obese commercially insured adults without diabetes. J Manag Care Spec Pharm. 2024;30(8):860–867. doi:10.18553/jmcp.2024.23332
3. Gasoyan H, et al. Changes in weight and glycemic control following obesity treatment with semaglutide or tirzepatide by discontinuation status. Obesity. 2025. doi:10.1002/oby.24331
4. Wilding JPH, et al. Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension. Diabetes Obes Metab. 2022;24(8):1553–1564. doi:10.1111/dom.14725
5. Rubino D, et al. Effect of continued weekly subcutaneous semaglutide vs placebo on weight loss maintenance in adults with overweight or obesity: the STEP 4 randomized clinical trial. JAMA. 2021;325(14):1414–1425. doi:10.1001/jama.2021.3224
6. Truveta Research. Real-world temporal and indication-specific variation in drivers of GLP-1 RA discontinuation. ISPOR 2025. https://www.truveta.com/blog/research/ispor-2025
7. Wilding JPH, et al. Once-weekly semaglutide in adults with overweight or obesity. N Engl J Med. 2021;384(11):989–1002. doi:10.1056/NEJMoa2032183
8. Klemets A, et al. Fecal microbiome predicts treatment response after the initiation of semaglutide or empagliflozin uptake. medRxiv. 2024. doi:10.1101/2024.07.19.24310611
9. Zeng M, et al. Crosstalk between glucagon-like peptide 1 and gut microbiota in metabolic diseases. Front Immunol. 2024;14. doi:10.3389/fimmu.2023.1123202
10. Gawron-Skarbek A, et al. Effects of GLP-1 Analogues and Agonists on the Gut Microbiota: A Systematic Review. Nutrients. 2025;17(8):1303. doi:10.3390/nu17081303
11. Heni M, et al. Heterogeneity in response to GLP-1 receptor agonists in type 2 diabetes in real-world clinical practice: insights from the DPV register. Lancet Public Health. 2023;8(4):e250–e260.
12. Dawed AY, Reddy V, et al. Pharmacogenomics of GLP-1 receptor agonist response. (Under active characterization; see: Identification of Novel Polygenic Drivers of GLP-1RA Treatment Heterogeneity, CEE Research, 2025)
13. Prime Therapeutics. GLP-1 Therapy to Treat Obesity Among Members Without Diabetes: Three-Year Persistence Analysis. 2024. https://www.primetherapeutics.com
14. Alnaser-Alkhalefa B, et al. Real-world titration, persistence and weight loss of semaglutide and tirzepatide in an academic obesity clinic. Obesity (Wiley). 2025. doi:10.1002/oby.[forthcoming]