A comparison of measures used to describe adherence to glaucoma medication in a randomised controlled trial

Understanding the magnitude of non-adherence in pre-marketing clinical trials and clinical practice is essential. However, accurately measuring non-adherence to medication is problematic, and the variety of adherence measures and/or calculation methods has led to highly variable results.

To compare commonly used methods for measuring adherence to eye drop therapy in order to assess which methods achieve the most complete datasets over an 8-month monitoring period, to quantify the magnitude of variance in adherence estimations and to review the different methods used to calculate or interpret adherence data.

Adherence was measured electronically for 8 months by participants administering eye drops using a Travalert® dosing aid. The mean number of Travalert dosing aid recorded doses administered over the monitoring period was used to calculate a percentage adherence score. In addition, the value of graphically presenting Travalert dosing aid data to classify patterns of adherence behaviour was explored. The validated Morisky Measure of Adherence Scale and questions requesting participants to report the Frequency of Missed Doses were two measures of self-reported adherence calculated for each participant. Finally, medication possession ratio was calculated from expected repeat prescription orders compared with actual repeat prescription orders.

For the 208 recruited participants, self-reported adherence was the most reliable method of collecting complete datasets over the 8-month period; 16% of self-reported adherence data were missing compared with 45% of Travalert dosing aid data missing at 8 months. The mean adherence measured over the monitoring period by the Travalert dosing aid was 77%. When adherence measures were dichotomised into adherent and non-adherent groups, the Travalert dosing aid found 54% of participants were adherent, compared to 60% Morisky Measure of Adherence Scale and 57% Frequency of Missed Dose self-report measures. However, there was poor agreement between the Travalert dosing aid measured adherence and self-report measures. Medication possession ratio was not a reliable measure of persistence with medication. Graphical Travalert dosing aid data presentation provided additional information about participant behaviour by indicating that most non-adherence was due to participants taking drug holidays rather than missing occasional doses.

The analysis provided evidence about the inconsistencies between different monitoring strategies and adherence measures. Furthermore, the analysis highlighted the difficulties in collecting complete data for studies investigating chronic, slowly progressive conditions that require long-term follow-up. Future adherence studies could use multiple methods for quantifying and classifying adherence in parallel, both to maximise precision of adherence estimates and to facilitate comparison between studies. However, the authors are cautious of the effect of using multiple adherence measures on participant behaviour and their potential reactivity effects is an area for future research.