The Medication Event Monitoring System ( MEMS^®) uses an electronic pill bottle top to compile dosing histories of ambulatory patient prescribed oral medication, and is the gold standard for measuring patient concordance [1]. While the MEMS^® is a valuable tool for identifying poor concordance it is relatively expensive (£50 each container) and therefore may not be feasible to use for each drug in individual patients for the whole battery life of 3 years. The aims of this study were i) to examine the relation between warfarin 1-month dosing history and INR monitoring over the preceding 12 months in patients on longterm therapy and ii) to investigate the variation in concordance measured by MEMS^® over 4 months.

Forty patients were recruited from clinics at the Royal Hallamshire Hospital and monitored for 1 (n = 25) or 4 months (n = 15). The study had local ethics committee approval. Patients were dispensed their usual warfarin prescription in the MEMS^® container with additional blister packs if needed. The containers were re-used after a cleaning process approved by the hospital pharmacy. MEMS^® data were analyzed with its own Powerview software (Aardex Ltd, Switzerland). Concordance was assessed as the percentage of times the device was opened vs prescription instructions. Univariate and multivariate regression was used to examine the relation between warfarin MEMS^® data and INR monitoring. anova was used to examine concordance over the first month with that of the subsequent 3 months in patients monitored for 4 months.

Three patients withdrew from the study (n = 1, 1 month; n = 2, 4 months). Of the remaining 37 patients, mean age was 71.5, SD ± 7.7 years, 22 (59.5%) were male, the most common indication for warfarin was atrial fibrillation in 27 (73%), duration of treatment was 5.6 ± 3.3 years, daily warfarin dose was 3.5 ± 1.5 mg, and patients were on 5.0 ± 3.5 other drugs. INR tests in the preceding year were in target range for 63.3 ± 20% of the time, with the mean INR value 2.8 ± 0.6. MEMS^® measured concordance was very high (99.7 ± 4.9%, n = 37). There was no significant relation between 1 month concordance and INR monitoring. 1 month concordance explained < 1% of the total variance of INR tests in target range over the preceding year. Age, sex, duration of warfarin treatment, and number of other drugs taken, together accounted for 4.5% of the total variance. In the 13 patients with 4 months of MEMS^® monitoring, mean monthly concordance was 101.4 ± 6.1%, 99.7  ± 2.3%, 99.5 ±  2.0%, 102.6 ± 5.4% ( anova, p = 0.18).

Concordance with warfarin therapy measured for 1 month was unexpectedly high in patients on longterm treatment attending medical and anticoagulant clinics in Sheffield. In these patients there was no relation between concordance and INR monitoring over the previous year. Monitoring with MEMS^® for 1 month was representative of more prolonged monitoring for 4 months, and may provide a valuable measure of concordance with drug treatment. However it does not appear to be of value in patients with relatively stable INR values.