.jpg){kind=link}
- Published
- Updated
Baseline is more than a column label. The rule selecting a baseline observation affects change scores and pretreatment comparisons. Define both the eligible time window and the selection rule before analysis begins.
Define the question and assumptions
Distinguish screening, randomisation and first administration times. Where several measurements qualify, explain why the last value, an average or another rule answers the question. Design timestamp and data checks that prevent a post-treatment measurement from being selected inadvertently as the baseline observation.
Hypothetical study example
In a hypothetical trial, screening and randomisation measurements occur on different days. One site enters a repeat taken after the first administration into the baseline field. The team checks the source, applies the protocol window and tests whether the analysis derivation makes the same mistake elsewhere.
Design review checks
- Check the relationship between measurement and treatment using dates and times where necessary.
- Specify the selection order when multiple observations fall within the permitted window.
- Prespecify how analysis proceeds when no eligible baseline measurement is available.
Record the decision and its limits
Acceptance examples should include boundary dates, missing times and repeated measurements. An independent reviewer should be able to reproduce the selection algorithm. Do not redefine baseline after examining outcomes to produce more favourable changes; explain any justified change and its consequences in reporting.
Frequently asked questions
- Is the screening value always baseline?
- No. The protocol and analysis definition determine the eligible window and observation.
- Is averaging two measurements automatically better?
- No. The averaging rule needs justification from the clinical question and measurement properties.