Outcome Measure Considerations for Clinical Trials Reporting on ClinicalTrials.
gov
What is an Outcome Measure?
An outcome measure is the result of a treatment or intervention that is used to objectively determine the baseline function
of a patient at the beginning of the clinical trial. Once the treatment or intervention has commenced, the same instrument
can be used to determine progress and efficacy. Outcome measures normally stem from overarching goals and aims.
Outcome measures should be measurable, which indicates that they are assessed by a numerical value.
ClinicalTrials.gov Outcome Measures Classification
One way to classify your outcome measures is distinguishing into three groups: primary, secondary, and exploratory.
Primary and secondary outcomes are required by law to be analyzed and reported in ClinicalTrials.gov if any data was
collected for the outcome. The primary and secondary endpoints should be pre-specified, meaning they are determined
before the start of the trial. We recommend that you do not include exploratory outcome measures as they are optional.
a. Primary Outcome Measure: defined by ClinicalTrials.gov as “the outcome measure(s) of greatest importance
specified in the protocol, usually the one(s) used in the power calculation. Most clinical studies have one primary
outcome measure, but a clinical study may have more than one”. The Primary outcome measures are the main
reason why you are conducting your study.
b. Secondary Outcome Measure: defined by ClinicalTrials.gov as “an outcome measure that is of lesser
importance than a primary outcome measure but is part of a pre-specified analysis plan for evaluating the effects
of the intervention or interventions under investigation in a clinical study and is not specified as an exploratory or
other measure. A clinical study may have more than one secondary outcome measure”.
c. Not required to report results on ClinicalTrials.gov:
a. Other Pre-Specified Outcome Measure: defined by ClinicalTrials.gov as “any other measurements,
excluding post-hoc measures, that will be used to evaluate the intervention(s) or, for observational
studies, that are a focus of the study”.
Exploratory Endpoints fall into this category
b. Post-Hoc Outcome Measures refers to outcomes that are specified AFTER the trial has started.
Considerations When Selecting Study Outcomes
1. Number of Primary/Secondary Outcome Measures: Listing a large number of outcome measures may
increase the chance of encountering issues when fulfilling the Clinicaltrials.gov reporting requirements. All primary
and secondary outcomes should have complete and accurate data when possible. Even if the endpoints are later
deemed to not be clinically relevant or only have limited data, they must still be analyzed and reported on if ANY
data is collected for that outcome. Consider outcomes that are clinically relevant, achievable, and address
realistic research questions.
2. Secondary vs Exploratory: Consideration should be given to whether a secondary endpoint may be better
identified as an exploratory endpoint. Appropriate secondary endpoints often are used to demonstrate additional
effects after success on the primary endpoint or to provide evidence that a particular mechanism underlies a
demonstrated clinical effect. If an outcome is only being used to frame future research or explore new
hypotheses, it may be better classified as exploratory. Exploratory endpoints may also include clinically important
events that are expected to occur too infrequently to show a treatment effect.
Version 11JUL2018 1
Funded by the NIH National Center for Advancing Translational Sciences through its Clinical and Translational Science
Awards Program, grant number UL1TR002541.
3. Outcome Measure Description: What information is being collected and how it is being collected needs to be
pre-specified for all primary and secondary outcomes. If a scale, grading method, device, etc. is being used to
evaluate an outcome, it should be pre-specified. Verbs frequently used to describe aims should not be used to
describe outcome measures such as “to determine”, “to assess”, and “to validate”.
Example 1: Instead of having the outcome be ‘to evaluate response rate’, it should instead be something
like ‘response assessed using Response Evaluation Criteria for Adverse Events (CTCAE 4). Response is
evaluated with the use of MRI or CT scan.’
Example 2: Instead of “measure of blood pressure”, it should be instead “change in systolic blood
pressure from baseline at week 12”.
4. Outcome Measure Time Frame: The specific time point(s) and overall duration of evaluation must be specified
in this section. If the outcome’s overall time duration can only be represented qualitatively (e.g. until the time of
disease progression), a specific numeric measure should then be included at the time of results reporting. This
could be something like the median duration of follow-up or the range of follow-up times.
Example 1: In the protocol, you may specify that response is evaluated after the end of every 28-day
chemotherapy cycle (± 7 days), until the time of disease progression, for up to two years. You may report
this in the time frame field on CT.gov as ‘baseline, end of every two 28-day cycles, up until disease
progression (maximum of two years)’. If there was not a specific endpoint cap you may instead say
‘baseline, end of every two 28-day cycles, up until disease progression, median duration of follow-up of
18 months’.
Example 2: Time frame should specify at which time points the outcome measure data were collected:
“change in systolic blood pressure was calculated at Week 12 minus baseline”.
5. Appropriate Quantitative Parameter for Each Outcome: Outcome measures cannot be reported as free text,
graphs, or qualitative information. It is allowed for the outcome measure to involve qualitative information, but it
must also be able to be reported in a quantitative/numeric fashion. Bear in mind that outcome measures that you
might normally report as qualitative (eg. graphs, images) for publishing purposes will need to be transformed into
quantitative data for results reporting in ClinicalTrials.gov. Each quantitative outcome measure also needs to
include an appropriate measure of dispersion.
Below are screenshots that show all the drop-down menu options for each section listed in the Outcome
Measure Data Table:
No measure of dispersion is needed for ‘Count of participant’ or other count data. For mean, median, least squares mean,
geometric mean, and geometric LSM, dispersion may include:
Version 11JUL2018 2
Funded by the NIH National Center for Advancing Translational Sciences through its Clinical and Translational Science
Awards Program, grant number UL1TR002541.
For the measure type of number, the measure of dispersion should be a confidence interval:
Statistical Considerations for Outcome Measures
The mathematical properties of an outcome measure determine the methodologies and conventions used to analyze and
report the results of a study. This section gives a brief summary of these conventions for common types of
measurements. Conventions for analysis and reporting for different type of measurements can have important
implications for ClinicalTrials.gov reporting so investigators should think carefully about their measures and build specific
and well-tailored analysis plans accordingly.
Continuous outcomes such as blood pressure, blood glucose, T-cell count and the like. Continuous outcome measures
are easily reportable to Clinicaltrials.gov but there are several considerations to note.
- They are typically reported as a measure of central tendency (mean, median, geometric mean etc…). The
measure of central tendency can be of values at a discrete time point (“the median pain score for the treatment
group was 9”) or of a change over time (“on average the treatment group saw a 3-point reduction in pain between
baseline and follow-up). An investigator should choose the correct measure of central tendency to report based
on the distribution of the data. Generally, means are reported for normal or near-normally distributed data and
medians or geometric means are reported for skewed data.
- ClinicalTrials.gov requires that investigators include not only measures of central tendency but also a measure of
the spread of data in your experiment such as the standard deviations, confidence intervals or interquartile
ranges. Standard deviations or confidence intervals should be included when reporting means and interquartile
ranges should be included when reporting medians. When reporting confidence intervals, the investigator may
need to refer back to any power analysis done prior to the study as this will inform which confidence interval
(95%, 90% etc.) would be most appropriate. Often an investigator will find that they did not include measures of
spread in their requests to data analysts. Once they are informed by Clincialtrials.gov that this is required it can
Version 11JUL2018 3
Funded by the NIH National Center for Advancing Translational Sciences through its Clinical and Translational Science
Awards Program, grant number UL1TR002541.
take considerable time to request this information again.
- Again, graphical depictions of data are not accepted by ClinicalTrials.gov so it is important to consider this upfront
and include as much quantitative information as possible. Some other measures to consider reporting are
measure of skewedness, kurtosis and tests for normality such as the Shapiro-Wilk test.
Ordinal outcomes such pain scales, patient satisfaction scales and other Likert-type measures are widely used in clinical
research. The considerations for reporting these types of measures are similar for that of continuous data. However, there
are some things to keep in mind.
- Clinical scales are often condensed as a summary measure. Sometimes as a mean or sum of a series of
questions or transformed to a dichotomous categorical variable. It is important for investigators to be familiar with
the conventions for analysis of clinical scales in the literature and be explicit in their definition of outcome
measures.
- Clinical scales are sometimes limited in range (consider a 1-5 scale) which can make their analysis as continuous
measures tricky. In these cases, investigator should consider alternative methods of analysis such as
dichotomization (dummy coding).
Categorical outcomes such as death deserve special consideration. Investigators should be explicit on how these
outcomes will be presented.
- The most common presentation of categorical outcomes are proportions, but categorical outcomes may also be
presented as counts, or relative measures such as risk ratios and odds ratios. It is important for the integrity of the
experiment’s design and Clincialtrials.gov reporting that investigators are very specific when registering their trial
in how these outcomes will be measured and presented.
- Measures of dispersion should also be included when reporting proportions or relative measures of effect. Much
like the presentation of means, proportions and relative measures of effect should be presented with confidence
interval. Again, look to any power analysis that was done for the study to decide which confidence interval level
should be used, even though the convention is 95%.
- In certain clinical trials, investigators may wish to follow an outcome such as death out to a time-point years away
from the start of the trial. Investigators should consider the feasibility of these timelines and this is a good example
of how outcome measures can frame the timeline of an experiment. There are methods available to investigators
for dealing with loss to follow-up in longitudinal studies such as reporting relative measures as rate differences or
rate ratios with person-time used for denominators. As noted before Clinicaltrials.gov does not accept graphical
depictions of data. This can present issues for reporting results for survival analysis which commonly feature a
survival curve. Investigators should keep this in mind if a survival analysis is planned and they should consider
reporting cumulative probabilities of death during a time period or median survival time. Investigators can also
consider calculating a hazard ratio for survival analysis.
Version 11JUL2018 4
Funded by the NIH National Center for Advancing Translational Sciences through its Clinical and Translational Science
Awards Program, grant number UL1TR002541.
Examples:
Version 11JUL2018 5
Funded by the NIH National Center for Advancing Translational Sciences through its Clinical and Translational Science
Awards Program, grant number UL1TR002541.
