The stress study used within-person questionnaire estimates, carried-forward observations and exclusions for incomplete data. Calling the entire analysis simply complete or imputed loses important distinctions.
Items, visits and outcomes differed
For isolated missing questionnaire items, the authors estimated an average scale or subscale score for that participant. Some missing repeated measurements were replaced using the last observation carried forward. Other incomplete outcome series were excluded when the authors considered transposition unrealistic.
The exclusions differed by outcome: the report describes incomplete data affecting cortisol, state anxiety and negative self-statements in different participants. A single enrollment denominator therefore cannot automatically be attached to every reported comparison.
A replacement is an analytical assumption
Carrying a value forward assumes that the earlier observation can stand in for an unobserved later value. That is different from obtaining another blood sample or questionnaire. Within-person item substitution makes another assumption about the missing response. Neither method identifies why data were absent.
What a defensible summary retains
Report the outcome, its usable sample and its missing-data method together. The study’s nonsignificant direct CBD-placebo contrasts should remain visible; missingness alone proves neither hidden benefit nor no possible benefit. The appropriate conclusion is limited evidence from a small laboratory experiment, with uncertainty that differs across measurements.
Related reading: CBD public-speaking research.
Research cited
Appiah-Kusi and colleagues (2020). Effects of short-term cannabidiol treatment on response to social stress in subjects at clinical high risk of developing psychosis. DOI: 10.1007/s00213-019-05442-6.
Leave a Reply