Dr Yogesh Gupta, Founder & CEO, Gaston AI
We have probably reached an unusual moment in healthcare: we are beginning to personalise the medicine before we learn how to personalise everything that happens around it.
GLP-1 therapies make that contradiction particularly visible. Their impact on obesity and metabolic disease is significant because they do something that medicine has struggled to achieve at scale for decades. In the United States, the share of adults taking GLP-1 for weight loss has almost quadrupled in two years, to 11%. In the UK, an estimated 1.6 million adults used a weight-loss drug in a single year.
Yet, once the prescription is written, much of the surrounding care still looks remarkably familiar. Eat Well. Get enough protein. Exercise. Come back for a review.
There is nothing inherently wrong with that advice. The problem is that the physiology of the person receiving it may no longer be static. That, I believe, is where more interesting GLP-1 conversation begins.
We are Grading New Medicine with an Old Ruler
The headline metric is still kilograms lost. But ten kilograms lost says little about what was lost: muscle or fat, nutrition adequacy, cardiovascular risk, physical function or whether the eating habits formed during treatment will last.
This does not mean that GLP-1 treatment inevitably leads to poor nutritional or
body-composition outcomes. It means something more fundamental: weight alone is becoming an insufficient description of treatment success.
The evidence deserves honesty rather than alarm. In the pivotal tirzepatide trial, about a quarter of the weight lost was lean mass, the same proportion as with placebo, and a 2026 study found strength largely preserved. Averages are reassuring. Patients are not averages, and a frail 75-year-old with little muscle to spare is not a typical trial participant.
What strikes me is the inconsistency. The FDA asks drug developers to measure body composition. Industry is now engineering it: in a phase 2 trial, adding bimagrumab to semaglutide raised the share of weight lost as fat from 76% to 92%. Yet in routine care, the instrument of record is still the scale.
I believe, the question needs to move from how much weight did we lose? to what kind of health did we create while losing it?
Continuous Medicine, Episodic Care
A person on GLP-1 is expectedly consuming considerably less. But if somebody is eating less, their protein adequacy, micronutrient density and food quality can no longer be treated as background considerations. If their biomarkers improve, deteriorate or change direction, the most appropriate dietary priority should change with them. If muscle or physical performance begins to decline, nutrition and resistance exercise must also be altered too.
However, unfortunately healthcare is not designed to operate this way. The current care model for a person who is on GLP-1 is episodic. GLP-1 acts on their physiology every hour of every day. Healthcare observes it occasionally, a consultation, a blood test, a weigh-in, a review months later. The result is a growing gap between what the treatment is doing and what the health system knows about what treatment is doing.
For example, In 2012, David Asch and colleagues observed that a person with chronic illness may spend a few hours a year with a clinician and 5,000 waking hours doing everything else. GLP-1 therapy makes those hours decisive. When appetite falls sharply, every meal carries more nutritional weight.
Furthermore, the early data are sobering. Among 461,382 adults with type 2 diabetes starting a GLP-1, 22% were diagnosed with a nutritional deficiency within a year. Fewer than one in ten had seen a dietitian in the six months before starting. In a US cohort of 125,474 adults, 65% of those without diabetes stopped treatment within a year. When the newer drugs stop, an Oxford meta-analysis projects that weight returns to baseline in about a year and a half.
From Personalised to Treatment-responsive Nutrition
Healthcare already has an abundance of data and dashboards. In the PREDICT study, people eating identical meals showed markedly different glucose and fat responses. Imagine if we combine clinical history, biomarkers, medication, dietary behaviours and relevant lifestyle information and ask a very simple set of questions: what has changed? What matters NOW because of these changes? And what, if anything, should this person do differently?
For years, the industry has equated personalisation with measurements: more biomarkers, more sensors, more microbiome data, more scores. The real threshold for precision nutrition is not whether we can describe an individual in extraordinary details. It is whether the information changes the decision in a way that is useful, evidence-based and relevant in real-time to that individual.
As treatment reshapes appetite, glucose regulation and body composition, the person at month nine may need different guidance from the person who started at month one.
Trajectories also diverge widely. In the tirzepatide trial, 57% on the top dose lost at least a fifth of their weight; about one in ten lost less than 5%.
What this demands is treatment-responsive nutrition: guidance that adapts as the treatment changes the physiology of the person receiving it. A falling ferritin, a fortnight of
low protein intake or a weakening grip should change the plan within days, not at the next review.
Our work at Gaston AI has firmly pushed us in this direction. In thinking about nutrition around GLP-1 treatment, for example, the interesting problem is not generating another generic meal plan with somebody’s name attached to it. It is understanding how biological risk, changing biomarkers, nutritional requirements and real food choices can be connected into an adaptive decision system.
The Risky Shadow System
Inevitably, when a formal system leaves gaps, an informal one fills them. In May 2026, near Northampton, the MHRA made its largest ever seizure of unlicensed weight-loss medicines. Among the roughly 12,000 doses was retatrutide, which its developer does not expect to submit for approval until 2027. The illicit supply chain is running ahead of the regulator. In the USA, “GLP-1 friendly” has no regulatory definition, yet the USDA has received 21 requests to print such claims on meat and poultry.
This is why Gaston AI is working with the TRACE initiative. TRACE began as an evidence and provenance architecture in the petroleum sector and is now being adapted with Gaston AI for medicines, supplements and health products. The sectors could hardly appear more different, yet the underlying problem is surprisingly similar: how do you move from being told something is authentic to having evidence that allows it to be verified?
That matters enormously in the world of rapidly growing demand for GLP-1 medicines, supplements and increasingly personalised health products. This has also changed how I think about the word precision which has two dimensions. The first is relevance: is this intervention right for this person? The second is, reliability: can we trust the evidence, product and information behind that decision? Healthcare will increasingly need both.
The Breakthrough is the System
I suspect the significance of GLP-1 will eventually extend far beyond weight management. This powerful therapy has exposed weaknesses elsewhere in a system. For example, patients may get the prescription from one service, blood tests from another and nutrition advice, if any, from a third one. Fundamental to health, food and supplement choices happen largely outside healthcare.
The fact is, human physiology does not recognise the boundaries between medicine, food, biomarkers, behaviour and everyday life. That is the healthcare problem I increasingly find myself interested in and is the problem we are solving at Gaston AI.
Picture a 48-year-old woman starting a GLP-1 this week. She will eat less than she has in years, and scale will reward her. Her muscle, iron and blood results may be changing too. She would also constantly be feeling out of energy in the second half of the day. Unless something goes wrong, no one may check for months. She would continue following the generic advice throughout her treatment period:
Eat well. Get enough protein. Come back for a review.
What we now need is an intelligence layer capable of recognising her risks which have changed before they turn into clinical diseases. That a treatment has changed her nutritional requirements. That her biomarkers are moving in the wrong direction. That the food on her plate matches her biological and metabolic responses.
As interventions become more powerful, I suspect the next breakthrough will increasingly be the system around the intervention: one capable of measuring the quality of outcomes rather than a single headline number; following the patient between encounters rather than only during them; adapting as biology changes rather than assuming it is static; and verifying the evidence on which its decisions depend.
The challenge now is to build a health system capable of responding with intelligence. Precision health is not about knowing more. It is about knowing what matters now and acting before biology becomes disease
About the Author
Dr Yogesh Gupta is Founder and CEO of Gaston AI, a preventive health technology company working at the intersection of artificial intelligence, biomarkers, nutrition and preventive healthcare. His work focuses on how biomarker-led AI can help healthcare to translate early chronic disease risks into personalised, adaptive and evidence-led daily food decisions and prevent them before they manifest.
About Gaston AI
Gaston AI is building an intelligence layer for precision health, connecting biomarkers, clinical risk, nutrition and real-world choices to enable more personalised and adaptive interventions.
Its NutriIQ™ AI translates an individual’s changing biology and nutritional needs into actionable food intelligence, while its wider technology portfolio supports earlier risk identification and precision prevention. Gaston AI is also working with the TRACE initiative to bring verifiable provenance and evidence integrity into medicines, supplements and health products.
Together, these capabilities address two essential dimensions of precision health: relevance, determining what is right for an individual at a particular moment and reliability, ensuring the evidence behind that decision can be trusted.
Read more: What was Old is New Again: Strengthening Your GEO/AEO Presence with Public Relations
