Adherence By Any Measure Still Matters

Home>
Specialty Care>
HIV/AIDS>
Research Notes

Adherence By Any Measure Still Matters

No single method of measuring adherence seems adequate for accurately assessing this important variable.

Despite the central importance of medication adherence in HIV management, measuring it accurately continues to challenge researchers. Working with patients at an HIV clinic in Los Angeles, researchers compared 3 different ways of measuring adherence and devised a composite score that they suggest may be more accurate than any single measure alone.

For a year, 108 adults newly started on PI or NNRTI therapy were given their medication in Medication Event Monitoring System (MEMS)-capped bottles containing a computer chip, which records each time the bottle is opened. Once a month the MEMS data were downloaded, nurses performed pill counts (or checks of levels of liquid medication), and patients' viral loads were measured. Twice during the year (weeks 8 and 24) patients reported on their own medication compliance by estimating how many times during the previous week they had missed doses.

A "composite adherence score" (CAS) was then devised for each monthly visit based primarily on the MEMS-cap data, but using pill counts and interview data to supplement missing or defective cap data and to compensate for patients who used pill boxes to lay out a week's pills all at once.

Although adherence fluctuated substantially over the year, MEMS-cap data consistently provided the lowest estimate for any 4-week period, and interview data, when available, provided the highest estimate. Pill counts estimated adherence at higher levels than MEMS-cap data, and the CAS average was generally at a level between these two.

Each of the adherence measures was compared with patients' viral-load measurements at the end of each 4-week interval. The CAS estimate of adherence had the overall best sensitivity and specificity for predicting an undetectable viral load compared with other measures of adherence. Reported nonadherence in an interview, for instance, had an estimated sensitivity of 14% and a specificity of 96% for identifying true nonadherence.

These findings are directed primarily to researchers seeking precise measures of adherence -- for them the bottom line is that a marker incorporating several measurements may be more accurate than any one measure alone. For clinicians, the study reiterates the point that patients claiming to take their medication often do not, although self-reported nonadherence is likely to be accurate and should be taken seriously. However, 2 editorialists caution against giving up on the interview as a way to measure adherence: Refined and used properly, it remains the clinician's primary tool for identifying patients whose adherence may be improved.

— Abigail Zuger, MD

Published in AIDS Clinical Care July 1, 2001