DEFINITIONS
Bias: “Systematic tendency of any factors
associated with the design, conduct, analysis,
and evaluation of the results of a clinical trial to
make the estimate of a treatment effect deviate
from its real value,” including inappropriate
assignment of subjects to treatment/control
groups, protocol violations, and exclusion of
subjects from analysis based on knowledge of
study outcomes.2
Operational Bias: “Bias introduced through
deviations in conduct.”2 When information from
an ongoing trial causes changes to the sample,
trial conduct, or predefined outcomes that could
impact the conclusions of the trial.2
Randomization: To reduce bias, participants
are assigned to “treatment or control groups
using an element of chance.”1 This permits
generalizations about outcomes.3
Type I error: Stating the treatment is more
effective than the control or comparison, when it
is not (i.e., false positive).
Type II error: Stating the treatment is not more
effective than the control or comparison, when it
truly is (i.e., false negative).
Interim Analyses
Efficacy analysis: Pre-planned analysis conducted
prior to the end of the trial to determine if
the treatment effect has been established;4 in
this case, it might be unethical to continue to
randomize patients to the control arm.
Futility analysis: Pre-planned analysis conducted
prior to the end of the trial to determine if the
treatment effect is unlikely to be statistically
significant if the trial is continued to the end;4
in this case, it might be unethical to continue to
randomize patients to the active treatment arm.
Analysis Adjustments for Missing Data
Baseline Observation Carried Forward (BOCF):
A single-value imputation method to fill in for
missing outcome values that assumes that the
baseline value reflects the long-term values.
The baseline value is used in the imputation
of the missing values.3,5
6
Last Observation Carried Forward (LOCF): A singlevalue imputation method to fill in for missing
outcome values that assumes that outcome does
not change after dropout, or the last observed
value reflects the long-term value. The last
observed value is used in the imputation of the
missing values.3,5
Analysis Populations
Full Analysis Set: “The analysis set which is as
complete as possible and as close as possible
to the intention-to-treat ideal of including all
randomized subjects.”2 Analysis of this set
typically requires imputed values or modeling
for missing data.3,5
Intent-to-Treat Analysis: Analysis is conducted
based on the planned treatment regimen for a
subject rather than the actual treatment given.
To achieve this, participants must be followed up,
assessed, and analyzed as part of the intended
group regardless of their compliance.2
Modified Intent-to-Treat: Although the definition can
differ between trials, a modified intent-to-treat
analysis can often mean that some randomized
participants are excluded from the analysis.6
Some examples might be participants who did
not receive any investigational treatment or did
not receive a predetermined minimum amount
of the investigational treatment.
Per-Protocol Analysis: Analysis of participants
who “complied with the protocol sufficiently to
ensure that these data would be likely to exhibit
the effects of treatment.”2 Essentially, these
participants followed the protocol without any
major deviations.
Safety: To review the medical risk associated with
a product, analysis of participants who received
any investigational study treatment as part of the
trial to determine the treatment’s adverse event
profile.2 Subject summaries are based on the
treatment received rather than the randomized
treatment.