What the literature actually says.

In the largest study to measure it directly, 53.6% of US adults with overweight or obesity discontinued a dual-labeled GLP-1 within one year. Among patients without type 2 diabetes (the population a weight-management practice mostly treats) it was 64.8%. Among those with type 2 diabetes, 46.5%.

Discontinuation is not permanent for everyone: of those who stopped, 36.3% of patients without diabetes had reinitiated within a year. The panel does not simply shrink; it churns.

Rodriguez et al., JAMA Network Open 2025, n=125,474, initiations 2018–2023, electronic health records. Discontinuation defined as 60 or more days without any GLP-1 on hand.

Among patients who discontinued, the reasons recorded in clinical practice were: cost or insurance 47.6%, side-effect intolerance 14.6%, drug shortage 11.8%, unspecified 11.1%, other 10.8%, switched to a compounded product 2.4%, and unsatisfactory weight loss 1.7%.

The distribution is the honest frame for what software can and cannot do. Nearly half of discontinuation is financial, and no amount of well-timed messaging changes a denied prior authorization. About one in seven stopped over side effects they could not tolerate, and the same study found that patients stopping for cost tended to do so later in treatment, while those stopping for side effects did so earlier.

Gasoyan et al., Obesity 2025, n=288 patients who discontinued within a year, Cleveland Clinic health system.

One-year persistence among commercially insured adults without diabetes rose across successive cohorts: 33.2% of those starting in 2021, 34.1% in 2022, 40.2% in 2023, and 60.9% in the first half of 2024.

It is tempting for a company like ours to read that as proof that management works. The authors do not, and neither will we. They attribute the improvement primarily to the resolution of GLP-1 supply shortages, listing better dose-escalation protocols, side-effect management and care programs only as possible additional contributors. Part of the change is measurement rather than behavior: as compounded products left the market, patients who had been buying outside the pharmacy benefit reappeared in claims data.

Neither study contains a variable for whether a patient received support. Attributing the improvement to management would be citing a paper for a conclusion it explicitly subordinates.

Marshall et al., JMCP2026, n=33,607. Longer-horizon figures exist (the same insurer reports 1 in 12 still on therapy at three years), but that cohort started in 2021–early 2022, and the publisher states it “may not be reflective of the current state.” We do not use it as a present-tense claim.

There is no randomized trial showing that care management, structured follow-up or telehealth check-ins improve GLP-1 persistence in the United States. We looked for one specifically, because our product would be easier to sell if it existed.

The best-powered digital datasets do not make the case either. Two analyses of a direct-to-consumer weight service that includes health coaching reported 12% adherence at twelve months among 4,535 semaglutide patients, and 16.1% at twelve months among 4,309 tirzepatide patients. Definitions are stricter and the market is cash-pay, so these are not comparable to the insurer figures above but they are emphatically not evidence that coaching raises persistence. Findings that engaged patients persist longer are reverse causation: the patients still messaging their coach are the ones still on the drug.

So the claim we make is narrow and mechanical, not clinical. About one in seven patients who stop cite side effects they could not tolerate, and side-effect discontinuation happens early. That is the kind of thing a program running every day is built to catch. Whether catching it changes twelve-month persistence is an open question, and we intend to be the ones who answer it.

Detecting a ten-point difference in twelve-month persistence between two protocol versions (say 33% against 43%) at 5% significance and 80% power needs roughly 380 patients per arm. A program managing a few thousand enrolled patients across a modest number of clinics produces that comparison for several protocol variants at once, inside a year, with directional readings in a quarter.

Where a source is a company research brief rather than peer-reviewed work, it says so above. Prime Therapeutics is a pharmacy benefit manager owned by Blue Cross Blue Shield plans, which has its own interest in persistence findings.