The natural history of medication compliance in a drug trial: limitations of pill counts.

To assess medication compliance over time, we prospectively performed pill counts among 121 ambulatory hypertensive subjects for less than or equal to 12 months. Prescribed regimens consisted of pinacidil or hydralazine administered four times a day and of secondary drugs administered up to twice daily. Surreptitious pill counts occurred every 1 to 12 weeks. Among a middle-aged subject group that had been selected for high rates of compliance, we observed mean compliance rates that approximated 100%. We noted marked intrasubject and intersubject variability for any one medication, between medications, and over time. From baseline blood pressures (+/- SE) of 155.5 +/- 1.9/97.3 +/- 1.0 mm Hg, subsequent mean blood pressures varied by compliance subgroup: "hypocompliers" (less than 80%), 151.3/91.0 mm Hg; "hypercompliers" (greater than or equal to 120%), 147.6/91.4 mm Hg; and "eucompliers" (80% to 119%), 143.3/88.5 mm Hg (systolic blood pressure: F1,52 = -220.9, NS; diastolic blood pressure: F1,52 = -121.4, NS). We concluded that weekly pill counts indicated marked intersubject and intrasubject variability, obscured by long-term averages; that compliance lapses appeared to be random; and that excessive medication-taking was the most consistent with "pill dumping."