5 Treatments and Controls
Treatments are the groups representing distinct interventions that are applied to the subjects in an experiment.
Controls effectively help remove any known or unknown experimental effects when assessing the effects due to the treatment. Controls should be treated in the same way as treatment groups to avoid introducing bias.
The researcher must decide which control groups are required in the experimental design. The choice of control affects which conclusions may reasonably be drawn from the experiment.
It is often tempting to measure the size of the treatment effect by comparing post-treatment response to a pre-treatment baseline, as this comparison is within the same animal and so the animal itself appears to play the role of a control.
However, this kind of measurement does not control well between groups. Individual animal measurements can be subject to (potentially unaccounted-for) time-related and drift effects. Baseline responses may also vary between groups of animals, so to compare between groups a difference in difference analysis should be performed.
Introducing too many groups (treatment and controls) but not adjusting the total number of individuals can reduce sample size to a point where the experiment is unlikely to be informative.
Sample size calculations require you to know how many groups (control, treatment, etc.) you need before you perform the calculation.
5.1 Negative controls
Negative controls (e.g. no treatment applied) aim to ensure that an unknown variable is not influencing the experimental outcome, and to avoid false positives.
5.2 Positive controls
A positive control involves giving subjects a treatment that is known to have an effect. This is useful in that it can demonstrate that the experiment would be expected to be successful in terms of being able to detect a known effect (in comparison to a negative control), if there was one to be seen.
5.3 Comparative controls
A comparative control is a form of positive control that may involve giving subjects an alternative or best-in-class treatment. The difference between the compound under study and the alternative/best-in-class treatment can then be measured.
5.4 Vehicle controls
Vehicle controls are a form of negative control where, for example, if the active compound is given in solution, the solution (but no active compound) is given to the subject. This allows the researcher to identify the effect of the treatment with respect to the effect of the vehicle (e.g. induced inflammation).
5.5 Sham controls
Sham controls are similar to vehicle controls, but mimic surgical and other physical interventions, without application of the treatment.
Suppose you are running a experiment with animal subjects to test the effect of a novel compound and you find the following:
- The comparison between the novel compound and the vehicle control is not significant
- The comparison between the positive control and the vehicle control is significant
What can you conclude?
✗The experiment could not detect a treatment effect if there was one, but the novel compound does have an effect
✗The experiment could not detect a treatment effect if there was one, and the novel compound has no effect
✓The experiment would be able to detect a treatment effect if there was one, but the novel compound has no effect
✗The experiment would be able to detect a treatment effect if there was one, and the novel compound has an effect