Stefano Garzon

Research · Oct 23, 2026

Participant adherence: measurement, run-in and effect on the estimate

Concept

Adherence is the degree to which a participant follows the prescribed regimen. In drug trials it is often defined as intake of a predetermined fraction of the dose, for example 80%. Persistence is a different measure: the time a participant remains on treatment, regardless of the proportion of doses taken.

Definitions

A participant who stops study medication but keeps attending visits is not a withdrawal. That participant still contributes outcomes.

Effect on the estimate and on power

In an intention-to-treat analysis, low adherence dilutes the difference between groups. It underestimates both benefit and adverse effects. According to Friedman, to maintain power, a 20% reduction in adherence may require more than a 50% increase in sample size, and a 30% reduction may require doubling it. In noninferiority trials, dilution favors a conclusion of no difference (see noninferiority margin).

In COURAGE, 2,287 patients with stable coronary disease were randomized to PCI plus optimal medical therapy or medical therapy alone. According to Friedman, 46 of the 1,149 PCI patients never underwent the procedure, and 32.6% of the medical group were revascularized over a median of 4.6 years. The primary outcome did not differ (19.0% vs 18.5%; HR 1.05; 95% CI 0.87 to 1.27). The comparison becomes one between initial strategies.

Adherence is also associated with prognosis for reasons unrelated to treatment. Participants adherent to placebo have lower mortality (see CHARM and the healthy adherer effect). For this reason, analysis by level of adherence compares non-randomized groups.

Design factors

Short, supervised trials with simple regimens have better adherence. A review of studies using electronic monitoring showed an inverse relation between the number of daily doses and adherence. With four doses a day, on-schedule adherence was about 50%.

Run-in period

During a run-in, candidates receive placebo or active drug for weeks before randomization. Those who do not adhere or do not tolerate it are excluded.

In the Physicians’ Health Study, 33,223 eligible physicians entered an 18-week run-in. The 22,071 who had taken at least two thirds of their pills were randomized. Lang and colleagues estimated a 20 to 41% increase in adherence and a 34% reduction in sample size, with power equal to or higher than randomizing everyone.

An active run-in also excludes those with adverse effects. PARADIGM-HF had two consecutive run-ins, with enalapril and then sacubitril-valsartan. According to Friedman, about 20% of eligible patients were excluded, mostly for adverse events. The result answers the question for patients who tolerated both drugs.

Berger and colleagues showed that selection by run-in response changes the representation of subpopulations and affects external validity. Lee and colleagues compared 43 SSRI datasets in depression and found no difference in effect size between trials with and without a placebo run-in.

Measuring adherence

No single measure is complete.

Common errors when reading a paper

What to check when reading a trial

  1. Definition and method of measuring adherence.
  2. Adherence by group and over time, compared with that assumed in the design.
  3. Crossover rates in each direction.
  4. Presence of a run-in, its duration and the proportion excluded.
  5. Distinction between those who stopped medication and those who left follow-up.
  6. Use of concomitant therapies by group.

References