Extract age, sex, severity, and setting into fixed fields for each trial arm, then flag any value reported only for the total sample because it cannot be split by arm.
- Create one row per trial arm and one column per baseline field (age, sex, severity, setting). Enter values exactly as reported, keeping the reported format and any dispersion measure with its value. Expected result: a grid where each cell traces to a specific arm.
- Read the trial report and its registration record side by side. ClinicalTrials.gov defines structured registration and results fields for design, eligibility, and reported results, so use those fields to confirm arm labels and reported baseline data. Expected result: arm names and baseline values reconciled between the two documents.
- Mark each value that appears only for the combined sample, such as a total-sample mean age or overall sex distribution. Flag it in the row rather than assigning it to an arm. Expected result: every unsplittable value carries a visible flag.
- Decide how flagged values enter the table. Keep them in a separate total-sample column, or leave the arm cells empty with a note. Expected result: readers can see which values describe arms and which describe the whole trial.
- Check the remaining fields against the review's eligibility criteria, such as whether severity and setting are reported in comparable categories across trials. Expected result: a list of fields that need harmonizing before pooling.
Cochrane Handbook Chapter 10 covers missing data and sensitivity analyses, which helps when flagged values affect how a baseline characteristic is summarized across trials.