Benchmarking retention and adherence across therapeutic areas

An obesity brand plan and a lipid brand plan can both say "retention is strong" and mean numbers that differ by a factor of three. Roughly half of statin patients have stopped by month 12. In US commercial claims work on GLP-1s dispensed for weight loss, the share still on therapy at one year lands near a third, sometimes lower. Oral oncology often holds 80% or more of patients through a treatment course. Psychiatry splits in two: oral antipsychotics shed patients within months, long-acting injectables hold far longer. There is no "pharma average" worth benchmarking against, and brands that use one tend to find out in year two of the plan, when the refill revenue does not arrive.

The work here is choosing a defensible comparator, then rebuilding the plan around the number that comparator hands you.

Pick the comparator before you pick the metric

Three measures do most of the work, and the discipline is saying which one you mean:

• Adherence: how completely a patient covers the regimen while on it, usually PDC (proportion of days covered), the share of days in a window with drug on hand according to pharmacy claims.
• Persistence: how long the patient stays on therapy at all before discontinuing, in days or months.
• Switch-back rate: the share of patients who leave for a competitor or a generic and later return.

PDC ≥80% is the standard cutoff for "adherent," used by CMS Star Ratings and by pharmacy benefit managers. It is a claims artefact, not a clinical fact: a patient can hit 85% PDC with a cupboard full of unopened boxes.

The World Health Organization's 2003 report on adherence to long-term therapies put adherence to chronic medication in developed countries at about 50%. Broad and old, and still the right prior when you have nothing product-specific: assume half your patients are gone or under-dosing inside twelve months until claims say otherwise.

Comparator choice matters more than metric choice. Route of administration, whether the patient feels the disease, who initiates the refill, and whether coverage can be withdrawn mid-year will each split a therapeutic area into bands that share almost nothing.

Benchmark ranges by therapeutic area (US, directional, 2024-2025)

Directional estimates from published adherence literature and claims-based studies, not one brand's confidential data. Read them as ranges.

Sources for directional ranges: adherence reviews indexed via NCBI/PubMed and CMS Star Ratings technical notes. Re-verify against the specific drug class and the most recent claims cut before using any of this in a real benchmarking exercise.

Read the pattern: retention tracks felt consequence and support infrastructure far more than efficacy. Topicals do not underperform because they fail; patients stop because they succeeded.

Worked comparison: an obesity brand against the wrong band

A weight-management brand reports 12,000 first fills in a quarter and tracks pharmacy claims for 365 days. At day 365, 3,840 patients are still on therapy.

12-month persistence = Patients on therapy at day 365 / First-fill patients
                     = 3,840 / 12,000
                     = 32%

Compared against GLP-1 persistence in type 2 diabetes, 32% looks like a crisis. It is the wrong band. Novo Nordisk sells semaglutide in both indications, and persistence in the diabetes population runs higher than in weight loss for the same molecule: the diabetes patient has a durably covered chronic indication, an HbA1c reading, and a physician who treats stopping as a clinical event. The weight-loss patient has none of those three.

Against the weight-loss band, 32% is in range. So the conclusion is not "fix retention," it is "resize the plan." The average starter contributes something like five or six fills, not twelve. That number belongs in the LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → model the module builds elsewhere; the consequence here is arithmetic on the patient panel. At 32% annual persistence you need more than two new starts for every patient you want to still have next year, so a flat panel of 100,000 requires north of 200,000 starts across twelve months. Acquisition budgets sized on a twelve-month revenue assumption will not clear.

Two shocks to separate out before anyone judges the retention team. Supply: Novo Nordisk's Wegovy constraints in 2022-2023 held back new starts and interrupted patients already titrated. Coverage: several large employer and state employee plans dropped weight-loss GLP-1 coverage in 2024, and a patient whose benefit disappears in July did not churn, they were removed.

Switch-back rate: the metric most brands ignore

Raw discontinuation overstates true brand loss. A patient pushed onto a generic during a formulary exclusion (when a PBM removes a drug from its covered list) who returns when your brand is reinstated is friction, not attrition.

Switch-back rate = Patients returning to brand within 12 months /
                    Patients who discontinued brand

Figures cited informally in commercial analytics put this at 15-25% for brands with strong patient hub services. A high switch-back rate says your retention economics are healthier than the persistence curve looks, and it is the cleanest argument for funding hubs, which keep the on-ramp back open after a lapse. Psychiatry is the caution: a patient who stops an oral antipsychotic often stops in a relapse, and comes back through an emergency department rather than a pharmacy. Counting that return as recovered marketing value misreads what happened.

Reading these benchmarks in Europe

Adherence measurement is less standardised across the EU, because claims infrastructure is fragmented by country and much of primary care runs through public systems (NHS in the UK, statutory health insurance in Germany) rather than PBM claims databases. Persistence studies exist through national registries and bodies including the European Medicines Agency, but comparable PDC benchmarking is harder to source publicly and runs on a lag. Treat any EU adherence figure as a study-specific estimate rather than an operational benchmark, and say so out loud when you present it.

Turning benchmarks into marketing action

Once you know where you sit against a fair band, four levers move the number:

1. Reduce refill friction. 90-day fills and auto-refill enrolment lift PDC against 30-day fills, because every fill is a fresh decision point.
2. Size hub services to disease burden, not to brand ambition. Biologics and oncology brands with nurse onboarding and check-in calls beat their category baselines; a topical with a call centre mostly buys reporting.
3. Split churn by reason: cost, tolerability, formulary exit, and goal reached. Obesity makes the fourth category unavoidable, and a copay card does nothing for a patient who hit their target weight and decided they were finished.
4. Do not confuse measurement with behaviour. Otsuka's Abilify MyCite, approved by the FDA in 2017 as the first medicine with an ingestible sensor, gave clinicians confirmed-ingestion data in a category where adherence is the whole clinical problem. Commercial uptake stayed small. Knowing precisely when a patient stopped is not a program for keeping them.

Watch where your numbers come from. A customer data platformcustomer data platformSoftware that unifies customer data from every source into one persistent profile that marketing, sales and service teams can act on.View full definition → such as Segment (Twilio's, sold as exactly this identity-stitching layer) will tell you who opened the email and downloaded the copay card; it cannot see a pharmacy fill. Persistence comes from claims bought through a data vendor and arrives weeks late. Teams that let platform-visible engagement stand in for persistence report retention that does not exist, and which engagement signals actually predict prescribing is the question the sibling lesson on that link answers.

🎬 [VIDEO: "Medication Adherence: Why Patients Don't Take Their Medicine" - youtube.com - a clinician-oriented explainer on the drivers of non-adherence, useful context for why marketing levers only solve part of the problem]

Key Takeaways

• Never benchmark against "pharma average." Compare within disease severity, route of administration and indication. GLP-1 persistence in obesity and in type 2 diabetes are different numbers for the same molecule.
• The realistic 12-month spread runs from roughly a third (obesity, oral antipsychotics, topicals) to 70-85% (oral oncology, biologics), with lipids and oral diabetes in between. An order of magnitude of plan risk sits inside that spread.
• PDC ≥80% over 365 days is the standard adherence definition; confirm the measurement window before comparing figures across sources, and remember the WHO's ~50% prior for chronic therapy when you have no product data.
• Low persistence changes the plan, not just the retention tactics: at 32% annual persistence you need more than two starts per patient retained, which reprices every acquisition assumption.
• Switch-back rate separates real attrition from formulary and supply disruption, and is the strongest justification for hub investment, with psychiatry as the exception where a return often follows a relapse.
• In Europe, benchmarking is fragmented by national health system; cited figures are study-level estimates, not live operational benchmarks.