most drugs are eliminated from the body with
plasma half-lives of 12 h or less (Feely et al.,
1987; Pullaret al., 1991), such that the measure-
ment of a drug level in plasma reflects dosing only
during the past day or two. Moreover, a prominent
feature of patients’ dosing behavior is an improve-
ment in compliance in the day or two prior to a
scheduled examination the so-called ‘white-coat
compliance’, which is a considerable improvement
in compliance in the day or two prior to a scheduled
visit to the physician (Feinstein, 1990). Thus, for
most drugs, the measurement of drug concentra-
tion in plasma at the time of a scheduled examina-
tion reflects dosing only during a period of
frequently atypically good compliance. For those
few drugs that have exceptionally slow turnover,
however, periodic measurement of drug concentra-
tion in plasma gives a view comparable to that
provided by the low-dose, slow turnover markers
described next. One could, of course, contrive to do
unannounced sampling of blood, but it is very
costly, often impractical and intrusive. Certainly
the finding of zero concentration of drug in plasma
is clear evidence that the patient has not taken any
drug during a prior period of time equal to four
half-lives of the drug in plasma, for example, 48 h
for a drug with a plasma half-life of 12 h.
A useful but rather cumbersome direct method is
based on the use of a slow turnover chemical
marker substance incorporated into the drug
dosage form. With a low dose of phenobarbital
used as the marker, for example, the turnover is
slow enough so that a single measurement of mar-
ker concentration in plasma is indicative of dosing
over a period of a week or more. Marker methods
do not, however, indicate the actual timing of doses
but only aggregate dosing during a time window
defined by the pharmacokinetic turnover of the
marker. Another disadvantage, of course, is that a
series of blood samples are needed if one desires a
longitudinal record of aggregate drug intake over
an extended period of time.
Indirect
Prior to 1987, the indirect methods in use were
those that made it easy for patients to censor
evidence for delayed or omitted doses: history,
diaries and counts of returned dosage forms.
They have repeatedly been shown to be unreliable,
giving results that, as Pullaret al. (1989) put it,
‘grossly overestimate compliance’ in both trials
and practice. Other authors have drawn the same
conclusion in other kinds of study situations.
In 1987, the first electronic monitor for solid
dosage forms was introduced (Norell, 1984; Pullar
and Feely, 1990; Cramer, 1995; Urquhart, 1997).
This device monitors the opening of the drug pack-
age, by means of time-keeping microcircuitry that
registers the time and date of each opening and
closing of the package. With this system, the date
and hour at which the package is opened and
closed, as well as the interval between each pair
of opening/closing events during the whole obser-
vation period can be determined. Its only disadvan-
tage is that the actual intake of the drug cannot be
confirmed, except with a rather complex combina-
tion of pharmacokinetic projection of the time-
course of drug concentration in plasma with peri-
odic direct sampling to assess the reliability of
those projections (Urquhart, 1997). In addition,
experience has shown that patients not wishing to
take their medicine rarely go to the length of open-
ing the drug package at scheduled times, but not
taking the medicine, day in and day out, throughout
the whole course of prescribed treatment.27.5 Compliance during clinical
trialsAgrowingconsensus supportsthe measurement,by
reliable means, of compliance with the protocol-
specified regimen during clinical trials. This con-
sensus is reflected in comments by leading biosta-
tisticians – PaulMeier(1991),BradleyEfron(1991,
1998), Donald Rubin (1991, 1998) and Sir David
Cox (1998). The gold-standard method for evaluat-
ing patient compliance is now accepted to be elec-
tronic monitoring (Cramer, 1995; Kastrissios and
Blaschke, 1997; Urquhart, 1997), and the nature of
the data found with electronic monitoring is well
summarized in (Urquhart and de Klerk, 1998). The
difficultiesoftryingtointerpretdatafromunreliable
methods, for example, counting returned, unused27.5 COMPLIANCE DURING CLINICAL TRIALS 357