CS Wendel, Southern Arizona VA Health Care System; MJ Mohler, Southern Arizona VA Health Care System; K Kroesen, Southern Arizona VA Health Care System; NM Ampel, Southern Arizona VA Health Care System; AL Gifford, VA San Diego Healthcare System; SJ Coons, University of Arizona College of Pharmacy

Objectives: The validity of Medication Event Monitoring System (MEMS) data rests upon several assumptions, including: 1) subject took full dose each time bottle opened, 2) subject did not decant additional doses, 3) MEMS use did not alter adherence behavior, and 4) subjects consenting to cap are a representative sample. This study explored the practicality of meeting assumptions in measuring adherence to combination anti-retroviral therapy (HAART) in an HIV clinic.

Methods: We performed a cross-sectional study within the HIV clinic at the Tucson VAMC. HIV-infected veterans on HAART (stable regimen for at least 30 days) were eligible. A one-month prescription of one antiretroviral was dispensed with a MEMS cap. Exit interviews elicited self-reported adherence and qualitative impact of MEMS on adherence. Decanting behaviors before and during monitoring were tabulated. The primary index of MEMS adherence was the proportion of prescribed doses that were not missed. Measurement properties of other commonly used adherence indices were compared.

Results: Sixty-four male subjects completed MEMS cap monitoring. Twenty-seven (42%) had some decanting routine prior to monitoring. Twelve (44%) of these were daily decanters and 15 (56%) used weekly pill organizers. During monitoring, 10 (37%) did not stop decanting, 14 (52%) stopped decanting only the capped medication, and 3 (11%) stopped decanting all antiretrovirals. Certain other adherence strategies did not accommodate MEMS, such as signals involving loosening or inverting caps after taking a dose. Eight (13%) subjects said MEMS made adherence more difficult, whereas six (9%) said MEMS made them more conscious of adherence, and two (3%) mentioned both inhibition and reminding. Two subjects attributed skipped doses or time changes to the MEMS cap. All three subjects who altogether stopped using pill organizers reported that MEMS made adherence more difficult; their adherence was lower than that of the remaining group (69% vs. 89%, p=.07). The majority (71%) of subjects who refused to participate used pill organizers, suggesting potential selection bias. Theoretical inspection of commonly used adherence indices revealed several to have undesirable properties. Index weaknesses included 1) giving over- (or early) dosing equal impact as under- (or late) dosing, 2) giving partial-dose days equal impact as drug holidays, and 3) allowing extra doses on remote days to compensate for missed doses.

Conclusions: Personal adherence strategies that are incompatible with MEMS are common in patients on complex regimens. MEMS data on decanters may be biased, yet excluding decanters could further erode applicability of measures to all HAART patients. MEMS use may have affected adherence behavior. Multiple measures of adherence, including self-report, diaries, and decanting assessment are recommended. There is no standard set of adherence indices. Since the critical issue for HAART is continual maintenance of therapeutic action, indices that emphasize missed or delayed doses should be used.

Impact: Because it provides detailed, objective, and comprehensive data, MEMS is generally accepted as the best available technology for adherence measurement and has been employed in several VA-funded studies of HAART adherence. We recommend that adherence research methods within the VA be standardized by a consensus panel of experienced adherence investigators.