Trial adherence and real-world adherence are different quantities measured under different conditions, and the gap between them is one of the largest in modern pharmacotherapy. This reviews what pharmacy-claims and retrospective cohort analyses report about how long people actually stay on GLP-1 therapy, and what the studies identify as predictors.
Key Takeaways
- Real-world persistence at one year is substantially lower than adherence rates observed inside randomized trials.
- Claims-based analyses from pharmacy benefit managers and health-data organizations have repeatedly reported that a large share of patients starting GLP-1 therapy for weight management are no longer filling prescriptions at twelve months.
- Cost and coverage changes are among the most frequently identified discontinuation drivers in the real-world literature — a category that does not exist inside a trial, where drug is supplied free.
- Gastrointestinal adverse events cluster in the early weeks, which is why the discontinuation curve is front-loaded.
- Reaching a higher maintenance dose and achieving early weight response are both associated with longer persistence in observational analyses.
Why trial adherence overstates real-world persistence
A randomized trial supplies the drug at no cost, screens out participants with contraindications and competing conditions, schedules regular contact with study staff, and selects for people willing to enrol in a research protocol. Every one of those features raises adherence relative to ordinary care.
Real-world use removes all of them. The patient pays, or their coverage does until it changes. Titration support may be a portal message rather than a study visit. There is no protocol requiring a return visit. The result is a persistence curve that looks nothing like the completion rate of the registration trial.
What claims data measures — and what it misses
Most real-world persistence estimates come from pharmacy claims: a patient is counted as persistent while prescriptions continue to be filled within an allowable gap. This method has real strengths — large samples, no recall bias, unselected populations — and specific blind spots.
- Cash-pay and telehealth-supplied medication is largely invisible. A patient who switches from an insurance-covered brand product to a cash-pay compounded product appears in claims data as a discontinuation. They have not stopped treatment; they have left the dataset.
- Fills are not doses. A dispensed prescription is evidence of acquisition, not administration.
- Reason for stopping is rarely captured. Claims show that a fill stopped, not why. Reasons come from survey and chart-review studies with their own limitations.
The compounded-market blind spot is particularly consequential for the 2023–2026 period, during which a large share of GLP-1 use in the United States moved through cash-pay telehealth channels that never touch a pharmacy benefit claim.
| Discontinuation driver | Typical timing | Captured well in claims data? |
|---|---|---|
| Gastrointestinal adverse events | Early — first weeks to first months | No — reason is not recorded |
| Cost or coverage change | Any time; often at plan-year boundaries | Partially — a stop is visible, the cause is not |
| Supply constraint | Episodic, market-wide | Visible only as population-level gaps |
| Insufficient response | Typically after several months | No |
| Goal achieved, patient elects to stop | Later | No — indistinguishable from failure to persist |
| Switch to cash-pay compounded product | Any time | No — appears identical to discontinuation |
What is associated with staying on treatment
Across observational analyses, several associations recur. All are associations rather than demonstrated causes, and confounding by indication is a live concern throughout.
Reaching a maintenance dose. Patients who complete escalation to a maintenance dose show longer persistence than those who discontinue during titration. Whether this reflects the dose itself or simply that tolerant patients both escalate and persist is not resolvable from the data.
Early weight response. Meaningful loss in the first months is associated with continued treatment. Motivation is the obvious mechanism; the observation is also unsurprising.
Coexisting type 2 diabetes. Persistence is generally higher when the indication is diabetes rather than weight management alone, which plausibly reflects both coverage differences and how the patient frames the treatment's necessity.
Continuity of clinical contact. Structured follow-up is associated with better persistence in the broader chronic-disease adherence literature and is consistent with what is reported here.
Front-loading of the discontinuation curve
The shape of the survival curve matters as much as its endpoint. Discontinuation is heaviest in the early period and the curve flattens thereafter. This aligns with the adverse-event timeline: gastrointestinal effects are concentrated during initiation and escalation and generally attenuate with continued exposure.
The practical inference drawn in the clinical literature is that the intervention window that matters most is the first eight to twelve weeks. Slower escalation, explicit preparation for what early side effects feel like, and accessible clinical contact during titration are the levers with the most plausible impact, precisely because that is when the losses occur.
Frequently Asked Questions
What proportion of patients are still on GLP-1 therapy at one year?
Published claims-based analyses have reported figures well below half in weight-management populations, with meaningful variation by dataset, population and definition of persistence. The compounded and cash-pay market is largely absent from these datasets, which likely biases the estimates downward.
Does discontinuation mean treatment failure?
Not necessarily. Claims data cannot distinguish a patient who stopped because of intolerance from one who reached their goal and elected to stop, or one who switched to a channel the dataset does not observe.
Are persistence rates improving as pricing falls?
Cost is a repeatedly identified driver, so falling prices would be expected to help. Whether that has translated into measurably better persistence is a question the published datasets have not yet answered for the most recent period.
Does the compounded market change these numbers?
Substantially, and in a direction that is hard to quantify. Cash-pay compounded use is largely invisible to pharmacy-claims analysis, so both discontinuation and switching are systematically misclassified.
References
- Prime Therapeutics. Real-world analyses of GLP-1 persistence in weight-management populations.
- Blue Health Intelligence. Retrospective claims analyses of GLP-1 discontinuation.
- Gleason PP et al. Real-world persistence and adherence to glucagon-like peptide-1 receptor agonists. Journal of Managed Care & Specialty Pharmacy.
- Rodriguez PJ et al. Retrospective cohort analyses of GLP-1 discontinuation and reinitiation. JAMA Network Open.
Comparing providers, not just trials?
Our homepage keeps a maintained side-by-side grid of GLP-1 telehealth providers with current pricing, formats and disclosures. That is where the provider comparisons live — this research desk stays citation-first.
Open the provider comparison grid