Everybody quotes a persistence number, this site included. This review is about why two careful studies of the same population can produce different ones, and the answer is a parameter nobody prints next to the percentage — the layer above any single figure for how many people stop.
How stopping gets counted
Persistence in this literature is measured from pharmacy refills. [1] You are persistent until a gap opens between refills that is longer than some allowed threshold, and that threshold is a choice the analyst makes.
Allow a short gap and a person who was late by a few weeks is recorded as having stopped. Allow a long one and a person who genuinely quit stays counted as continuing for months. The review found estimates varying by exactly this, and by the analytic approach used around it.
What the numbers came to
Persistence generally declined over time and often fell below 60% by 12 to 24 months. In one large cohort discontinuation reached 64.1% at two years — which is 35.9% still on treatment, this desk’s subtraction, put on the same scale as the first figure so the two can be read together: under 60% against 35.9%, at one to two years and at two.
Persistence varied with age, income, gastrointestinal side effects, how much weight came off, body mass index measures, which agent, which formulation and the dosing schedule. Some of those are clinical. Income is not.
The part nobody studied
Only two of the seven studies reported patient-level reasons for stopping, most commonly adverse effects. None described any structured behavioral or supportive strategy.
That is a striking absence in a literature this size. The field can count who stopped several different ways and has barely asked anybody why — and what happens when somebody restarts is a clinical question sitting inside those refill gaps that nobody is measuring either.
What a buyer takes from it
When you meet a persistence figure, ask what gap it allowed. A site quoting “two thirds stop within a year” and a study reporting 64.1% at two years may be describing the same behavior with different rulers.
And note the population: adults with both obesity and type 2 diabetes, not a general weight-loss buyer. Income being among the predictors is the finding that connects to everything else here, and it points the same way as the study where most eligible patients declined on cost. What you pay each month is on the six-month view; whether you are still paying is the thing nobody prices.