ANCOVA (adjusting for baseline)
Compares groups on a continuous follow-up outcome while adjusting for the baseline value of that same outcome.
Clinical example
Systolic blood pressure at 12 weeks in two trial arms, adjusting for each patient's blood pressure at baseline.
When to use it
- The outcome is a continuous measurement taken at baseline and at follow-up.
- You compare groups, typically in a randomized trial.
- You want a more efficient comparison than testing the change from baseline.
Assumptions to check
As for linear regression
Linear relation with the baseline value, and residuals with constant spread. Check the residual plots.
Baseline measured before treatment
The baseline value must be taken before the intervention starts, so that the treatment cannot affect it.
What to report
- Adjusted mean difference between groups with 95% CI
- The baseline variable and any other variables in the model
- The exact p-value, not just "p < 0.05"
Common mistakes
Don't test baseline differences in an RCT
With randomization, any baseline difference is chance by definition. Choose adjustment variables in advance, based on how strongly they predict the outcome.
Prefer ANCOVA over change scores
Comparing change from baseline between groups is less efficient than adjusting for baseline, and is affected by regression to the mean.
Alternatives
| If your situation is… | Use |
|---|---|
| Adjusting for confounders rather than for baseline | Linear regression |
| Repeated measurements | Linear mixed model |
In your software
# R
fit <- lm(sbp_12wk ~ arm + sbp_baseline, data = trial)
summary(fit)
confint(fit)
UNIANOVA sbp_12wk BY arm WITH sbp_baseline
/EMMEANS=TABLES(arm) WITH(sbp_baseline=MEAN) COMPARE(arm)
/PRINT=PARAMETER
/DESIGN=sbp_baseline arm.