Clinical Research Design and Patient-Oriented Outcomes 357(McDowell 2006; Streiner & Norman 2014).
There are a number of ways in which an instru-
ment can be tested for reliability. All represent
the average of the correlations among all of the
items at a single administration of the instru-
ment. In addition, to account for day-to-day
variability in responses, an assessment of the
reproducibility of responses is made by admin-
istering it to the owners of dogs that have sta-
ble disease and no change in treatment on two
different occasions (test–retest reliability). Once
an instrument is shown to be measuring some-
thing reliably, the process of determining what
that something is begins.
Validating an instrument is a process by
which the degree of confidence that can be
placed on conclusions drawn about an ani-
mal based on their score from that instrument
is determined. If other validated instruments
designed to measure the same attribute exist,
the experimental instrument is administered
along with the existing one to see whether there
is a strong correlation between the two. When
no other instruments exist, developers must test
‘construct’ validity, which is evaluated when
the attribute being measured cannot be directly
observed. For example, chronic pain cannot be
‘seen,’ but behaviors can be observed which,
according to current theories about chronic pain
in companion dogs, result from it. For exam-
ple, based on a construct of ‘Chronic Pain,’ dogs
with osteoarthritis (OA) that score high on a
newly developed instrument differ from dogs
that score low on it in terms of attributes such
as the results of gait analysis, assessment of
quality of life, and activity monitoring. There
is no single experiment to ‘prove’ the validity
of the instrument, but multiple, well-designed,
hypothesis-driven studies build a body of evi-
dence that the instrument is measuring what is
intended.
Developing a valid and reliable assessment
instrument can take many years, requiring spe-
cific expertise and financial support. Some of
the owner assessment instruments presented
in the First Edition of this book have under-
gone continued development and validation,
with new ones joining the group (Table 43.1).
Deciding which instrument to use, depends on
what needs to be measured and how comfort-
able the investigator is with the amount of data
available to support the validity and reliability
of the measure. Some are designed to measure
pain, others mobility, or stiffness, or function.
On many occasions investigators use more than
one instrument in a study to capture data on
several of these domains. Regardless of which
measure is chosen, details of the study design
are critical to capturing quality data.Clinical trial design
Overview
Due to its ability to minimize bias, randomized
controlled clinical trials (RCTs) are the ‘gold
standard’ methodology to evaluate interven-
tions for efficacy. Whether the results of a clin-
ical trial are truly unbiased, however, depends
on many features of its design.Defining the study population
For dogs with cruciate ligament rupture, rela-
tively narrow inclusion and exclusion criteria
are needed to manage the extensive variabil-
ity that can occur with the disease. Criteria to
consider include acute versus chronic duration;
bilateral versus unilateral disease; partial ver-
sus complete ruptures; amount of joint insta-
bility; prior surgical history; presence of menis-
cal damage; severity and location of OA in the
stifle(s); OA in other joints; bodyweight and
body condition score; and age. Which criteria
are chosen ultimately depends on the interven-
tion being studied and outcome measures used.Defining the control group
Control group definition in dogs with cruciate
ligament rupture depends on the intervention
under study, which can vary widely from anal-
gesics, disease-modifying agents, and rehabili-
tation protocols, to surgical procedures. When
placebo controls are not possible, standard of
care control groups can often be defined.Assignment to intervention versus control
groupsSelection bias occurs when study subjects with
one or more influencing factors appear more
frequently in one study group than in another.