The GLP-1 Data Nobody Puts in the Deck: Adherence, Regain and What Payers Now Model

There are two GLP-1 stories running in parallel, and health-tech founders keep pitching against the wrong one.

The first is the trial story. In the landmark obesity and cardiometabolic programmes, semaglutide and tirzepatide produced weight reduction at a scale no oral pharmacotherapy has approached, and the cardiovascular outcome data moved the class from cosmetic to clinical in the eyes of guideline committees. That story is real and it is durable.

The second is the real-world story, and it is the one that actually sets reimbursement. In claims-based analyses across commercial populations, a large share of patients who start a GLP-1 for weight management are no longer on therapy at twelve months. Published estimates vary with the population and the definition of persistence, but the consistent finding across independent datasets is that real-world discontinuation is substantially higher than trial dropout, and that weight regain follows discontinuation in most patients because the underlying physiology does not change.

Those two facts, taken together, are the entire commercial context for anyone building in this space.

Why the gap exists, clinically

In a trial, patients are screened for motivation, titration is supervised, side effects are managed by a study team, and the drug is free. Outside a trial, none of that is true.

Four things drive discontinuation in ordinary care. Gastrointestinal side effects during titration, when nobody has coached the patient through dose escalation. Cost and coverage churn, when a plan changes its formulary mid-year or an employer drops the benefit. Supply interruption, which has eased but trained a cohort of patients to stop and restart. And the quiet one: absence of any care model at all. A prescription written in a fifteen-minute visit with no follow-up is not a treatment plan.

The last of those is where the technology opportunity sits, and it is being under-built relative to how much capital has gone into simply getting people the drug.

What payers are now modelling

Employers and plans have moved from asking "does this work" to asking "what is my three-year cost per member if I cover it." The inputs to that model are unforgiving.

They are pricing in a high annual net cost per treated member, an uptake rate that keeps surprising them on the upside, a discontinuation curve that means a meaningful fraction of spend produces no durable clinical benefit, and a member turnover rate that means the plan paying for weight loss is often not the plan that would eventually harvest the avoided cardiovascular event. That last point is not cynicism, it is actuarial reality, and it explains coverage behaviour better than any other single factor.

The result is the coverage architecture now spreading across the market: prior authorization tied to BMI and comorbidity thresholds, a requirement to enrol in a lifestyle or care-management programme, continuation criteria that check for a minimum weight response at a set number of weeks, and in a growing number of plans, outright exclusion of the weight-management indication while retaining the diabetes indication.

What that means for anyone building here

Three implications, and they cut against the way most decks in this category are written.

Persistence is the product, not the prescription. Telehealth companies that optimised for speed of first fill built a business on the least defensible part of the value chain. The differentiated asset is a care model that keeps a patient on therapy at the right dose, manages side effects before they cause a stop, and documents response. That is what a plan will pay for, because that is what converts spend into outcome.

Deprescribing and maintenance are an open category. There is almost no accepted clinical protocol for what happens after target weight, and regain after discontinuation is the norm. Whoever produces credible evidence for a maintenance strategy, whether that is dose reduction, intermittent therapy, or a structured behavioural and resistance-training programme, will own a conversation that every medical director is currently having without data.

Body composition matters more than the scale. A meaningful share of the weight lost on these agents is lean mass. In older patients, that is a functional risk, not a cosmetic footnote. Programmes that pair therapy with protein adequacy and resistance training, and that measure composition rather than weight, are making a clinical argument that will age well.

The honest read

This class of drugs is one of the most consequential therapeutic developments of the last two decades, and the real-world data does not undermine that. What it shows is that a highly effective molecule delivered through a weak care model produces mediocre population results at a very high price.

That is not a pharmacology problem. It is a delivery problem, which is to say, it is a health-technology problem. The companies that understand the distinction will still be here when the coverage rules tighten again.