What are Historical Controls?A historical control is where old data is used to compare with new data from new trials. Information is essentially “borrowed” from historical data. The old data is usually from patients with the same disorder and might come from a variety of sources, including:
- Older clinical trials,
- Medical chart data,
- A prospective study,
- A patient’s own medical history.
The technique is widely used in cases where there are ethical reasons not to use a control group. For example, in cancer trials it wouldn’t be ethical to give no treatment or a “sugar pill.” It is also used when “Treatment-as-usual” is outdated or ineffective. If historical data is used instead of a control group, every person in the trial will receive the experimental drug; The control group is created from historical data.
Advantages and Disadvantages
A historical control group should be chosen so that the trial’s endpoints are comparable. For example, if a researcher is studying how long life is extended with a particular drug, they would want to draw data from a trial that also reported how long life was extended for. Advantages of using carefully chosen historical data:
- Costs can be cut dramatically,
- More resources can be allocated to the experimental group, resulting in more accurate point estimates, increased statistical power, and lower type I error rates.
If the controls are not carefully chosen so that they are reasonably compatible with the experimental group, this can result in:
- Inaccurate (biased) point estimates,
- Decreased power or increased Type I error rates.
Additionally, the use of historical controls can result in large selection bias (Baker & Lideman, 2001).
The following types are relatively easy to use:
- Dynamic borrowing: a middle-ground between using no historical data for the control or 100% historical data: When the data is evenly matched, borrowing takes place, and when the data isn’t well-matched, no borrowing happens. One form of dynamic borrowing is test-then-pool, where you basically run statistical tests before the trial to figure out if your participants are equal enough to use historical data for the control.
- Single arm trials: consist only of an experimental group. Historical data provides all of the control data.
- Pooled: the experimental groups and control groups are the same size, but historical data is pooled (added) to the data from the control group.
Other, more complex ways to use historical data for controls include power priors and hierarchical modeling.
Baker, S. & Lindeman, K. (2001). Rethinking H.C. Biostatistics. 2,4. pp. 383-396.
Viele, K. et. al. (2014). Use of H.C. data for assessing treatment effects in clinical trials. Pharm Stat. 2014 Jan-Feb; 13(1): 41–54.
If you prefer an online interactive environment to learn R and statistics, this free R Tutorial by Datacamp is a great way to get started. If you're are somewhat comfortable with R and are interested in going deeper into Statistics, try this Statistics with R track.Comments? Need to post a correction? Please post on our Facebook page.