GLP-1 Retention Benchmarks: the 12-Week Cliff

The benchmark numbers, sourced and dated

The early cliff: Blue Health Intelligence (an independent Blues-affiliated analytics company, published via the Blue Cross Blue Shield Association in May 2024) analyzed about 170,000 commercially insured members prescribed weight-loss GLP-1s from 2014 through 2023 and found 58 percent discontinued before completing twelve weeks of continuous treatment, the duration associated with clinically meaningful benefit, and more than 30 percent stopped within the first four weeks.

The longer horizons, shortage-era: Prime Therapeutics (a large pharmacy benefit manager) followed commercially insured adults without diabetes who started weight-loss GLP-1s in 2021 and early 2022: about 32 percent were still on therapy at one year, 15 percent at two years, and 8 percent at three, in analyses published from 2023 through mid-2025. A separate study of health-system records by Truveta researchers, published in JAMA Network Open in January 2025 across roughly 125,000 adults, found about 65 percent of patients without type 2 diabetes discontinued within a year and 84 percent within two.

The improvement nobody quotes: the same Prime Therapeutics research line, peer-reviewed in the Journal of Managed Care & Specialty Pharmacy in March 2026, found one-year persistence nearly doubled across initiation cohorts as the shortage era ended: from about 33 percent for 2021 initiators to about 63 percent for early-2024 initiators. The scary numbers describe the era of stockouts, pharmacy roulette, and higher prices; the current era retains far better, and honest benchmarking uses both.

The quotable version, with its sources attached: 58% of weight-loss GLP-1 patients historically stopped before the 12-week mark (Blue Health Intelligence, 2024), about a third of 2021-2022 starters lasted a year (Prime Therapeutics), and one-year persistence roughly doubled to ~63% for early-2024 starters (JMCP, 2026). The cliff is real, front-loaded, and increasingly fixable.

Why patients actually stop

The published analyses and the operating experience point at the same clusters. Early exits concentrate in side-effect surprise (the first weeks are the roughest, and patients who expected effortless results leave when reality arrives), access friction (a fill that doesn't arrive on time is a quit event, which is why the shortage era's curves were so bad), and cost shock (a price that felt survivable once feels different by month three). Later exits add goal ambivalence, plateau frustration, and the simple entropy of any subscription.

None of these are exotic, and that is the business insight: most discontinuation is not a clinical verdict, it is an operational failure someone could have prevented. The improvement between the 2021 and 2024 cohorts tracks exactly the operational variables (supply reliability and price) that changed.

What the cliff does to program economics

Run any program model against these curves and the sensitivity is immediate: a program whose average patient stays three months is a customer-acquisition treadmill, while the same program at nine months is a business. Acquisition cost is paid back across the retention curve, so the difference between the historical curve and the 2024-cohort curve is frequently the difference between negative and positive unit economics at identical pricing.

This is why sophisticated operators treat the first four weeks as the product: the BHI finding that nearly a third of patients historically left in month one means the cheapest retention gains in the entire category live in onboarding, first-fill reliability, and week-one and week-two experience. Nothing you do at month six matters for the patient who left in week three.

The program-design levers that move the curve

The levers are operational, and they respect the clinical line: expectation-setting before the first dose (what the first weeks actually feel like, framed honestly rather than promotionally); proactive care-team check-ins timed to the hard weeks, with clinical questions routed to the licensed clinicians who alone adjust treatment; supply reliability treated as sacred (the fill arrives before the last one runs out, every time); price predictability (medication-included flat pricing removes the month-three cost shock that pass-through pricing invites); and respectful off-ramps (pause options and honest exits convert some quits into returns, and returns are real: the JAMA-published analysis found a meaningful share of discontinuers later restart).

What the levers are not: outcome promises, pressure to continue, or any business hand on clinical decisions. A patient for whom therapy is not working or not appropriate should stop; the levers above exist to remove the operational reasons patients stop when therapy IS working. Programs that respect that line retain better and sleep better.

• Onboarding that sets honest week-one expectations beats any winback flow
• First-fill speed and refill continuity are retention infrastructure, not logistics
• Care-team touchpoints in weeks 1-4, with clinical questions routed to clinicians
• Medication-included flat pricing removes the month-three cost shock
• Pause and return paths: discontinuation is often an interruption, not an ending

Using these benchmarks without fooling yourself

Three honesty rules. Benchmark against the right cohort: the 2021-2022 numbers describe a shortage; if your program launched in 2026, the ~63 percent one-year persistence of recent cohorts is the bar, not the 32 percent of the bad years. Measure persistence the way the studies do (continuous therapy with a defined allowable gap) rather than flattering proxies like 'active accounts.' And segment the cliff: your week-four survival rate is a different metric from your one-year rate, is fixed by different levers, and per the BHI data is where the historical damage concentrated.

These are insurance-claims and health-records populations, not cash-pay telehealth cohorts, so treat them as reference curves rather than your forecast; cash-pay dynamics differ in both directions (self-selected motivated patients, but no insurance subsidy). Published cash-pay benchmarks barely exist as of September 2026, which is itself worth knowing when a vendor quotes you one without a source.

Where EmbedCare fits

The retention levers above are what EmbedCare operates as the platform: onboarding and expectation-setting content built into the patient experience, care-team touchpoints with clinical routing to the licensed network, owned pharmacy supply so refill continuity is an architecture rather than an aspiration, and medication included in flat product rates on GLP-1 programs so month three costs what month one did. Retention outcomes vary by program and audience and nothing here is a promised curve; a demo shows the machinery, and your own cohort data will show the rest.