two-week, open-label course of treatment. At the
other extreme, a randomized controlled trial
(RCT), using a double-blind, placebo-controlled
protocol and a 12-month follow-up in several hun-
dred patients, could use a statistically validated
QOL instrument. The results of the latter would
probably inspire more confidence than the ‘infor-
mal’ scenario, all other things being equal. How-
ever, it does not mean that the answers given using
the informal method are wrong; it simply requires
an appreciation of the trade-offs involved in how
data are collected. Furthermore, the former method
might be of more use than the latter in exploratory
pharmaceconomic research conducted in the ear-
liest stages of drug development.
The RCT, while regarded as a gold standard in
much of drug development, offers real challenges
to the pharmacoeconomist. The RCT is costly,
time-consuming, and may not always be ethical
(12 months of placebo?). Some types of outcomes,
such as compliance, do not lend themselves to
double-blind designs because such designs mask
one of the effects being measured. RCTs generally
strive to maintain high levels of internal validity at
the risk of reducing external validity. Biases to
internal validity affect the accuracy of the results
of the study, as they apply to thosewho participated
in the study (e.g. patient selection bias, crossover
bias and errors in measurement of outcomes).
Biases to external validity affect how well the
results may be generalized to the public at large.
Obviously, the choice of study design must take
potential biases into account. These factors are
somewhat analogous for pharmacoeconomic and
traditional clinical research.Selecting a QOL instrument
It is always important to select an instrument
that has adequate reliability and validity. Although
many instruments have been published, many of
these have little supporting validation. Another
source of information include the Medical Out-
comes Trust(2001;http://www.outcomes-trust.org).
Some instruments, such as the MOS-SF-36, a gen-
eric QOL instrument, seem to be gaining popularity,
and it is tempting to routinely incorporate these into
clinical studies. Many experts in the field recom-
mend that both a disease-specific and generic instru-
mentshouldbeusedineachstudy,inordertocapture
the broadest QOL information. Yet, excess burden
on patients can defeat the accuracy and complete-
ness of what is collected. Generally, if resources or
patient burden threatens, then most experts would
argue for retention of a disease-specific instrument
when it is only possible to use a single measure.
Standard operating procedures and quality ana-
lysis should be a part of every study in which the
company invests money to collect end points, beTable 23.3 Points to consider: incorporating pharmacoeconomic measures in clinical trials
Document the pharmacoeconomic objectives, methodology and analysis plan within the study protocol.
Measure outcomes in the most appropriate and most disaggregate units. Categories can always be collapsed at a
future time, but is impossible to split out variables beyond their original units. The sources of process and
outcomes data may vary.
Clinical data may be captured from providers, patients and medical records.
Resource use data may be obtained from patient, administrative databases, providers or charts.
QOL data should come from the patient. In some cases (very young, very old, mentally unstable) patient proxies
are used, but the patient should be considered the optimal choice.
The study design can affect thetypesof outcomes that can be reliably collected, and themannerin which the
outcomes can be collected.
Study design affects several parts of the evaluation process:
Cost of evaluation
Time required to conduct the evaluation
Accuracy of the information gained
Complexity of administering the evaluation
Ease of defending subsequent decisions made, based upon the evaluation
298 CH23 PHARMACOECONOMICS: ECONOMIC AND HUMANISTIC OUTCOMES