Allowance for Baseline Variables Flashcards
What are baseline variables?
Potentially prognostic factors that are measured before someone is allocated to one of the trial arms
Give some examples of baseline variables?
Demographic information (age and sex)
Characteristics (height and weight)
Etc.
What are the three main reasons for collection of baseline values?
Description of the participants
Comparability of the allocation between the arms of the trial
Inclusion in statistical analysis
Why is it important to measure baseline variables pertaining to description of participants?
External validity
Why is it important to gather baseline variables pertaining to comparability of the trial arms?
There is random chance that post-randomisation imbalance in baselines will occur. Adding descriptions of baseline variability will add credibility to trial results i.e. encourage confidence that unadjusted outcome analyses are without significant bias
How does the size of the trial affect baseline variability?
Larger trials are less affected
Should baseline imbalance be a cause for adjustment in primary analysis?
Not unless stated in the statistical analysis plan. If randomisation has worked correctly, baseline imbalance is due to chance.
For a better understanding of true treatment effect adjustment for baseline analysis may be important in exploratory analyses.
If a potential baseline variable may be important prognostically, how is it best to deal with it?
Design of randomisation e.g. stratification or minimisation
Change of baseline variables is a common endpoint. Suggest two reasons why it is not ideal?
- Statistical analysis of change from baseline tends to over-adjust for any baseline imbalance
- This method is not as powerful including the baseline as a corvariate in linear regression (ANCOVA) - due to its smaller standard error.
Moreover if the correlation co-efficient is less than 0.5 (baseline and follow-up values) then analysis of change has a bigger standard error than ignoring the baseline all together.
