Uncontrolled studies and historical controls
- Study design
Comparing today's patients with past ones, or with themselves before treatment, tends to exaggerate the benefit of a new treatment.
What it is
In an uncontrolled before-after study an outcome is measured before and after an intervention in the same group; in a historically controlled study patients treated now are compared with patients from earlier years. Both are cheap and fast, but any change over time (better general care, new diagnostic criteria, less severe patients, regression to the mean) is wrongly credited to the intervention. That is why they tend to overestimate benefit.
How it is built
- Uncontrolled before-after: baseline measurement, intervention, repeat measurement in the same group.
- Historical control: group treated now versus records of patients from another period.
- Historical control from another source: published results from another centre.
- Minimum improvement: a concurrent control group measured at the same times.
Worked example
- A hospital adopts a sepsis protocol in 2025. Mortality in 2023: 90 of 300 (30 %); in 2025: 64 of 320 (20 %).
- Historical comparison: 20 − 30 = −10 percentage points; RR = 0.20 / 0.30 = 0.67.
- A nearby hospital without the protocol went from 29 % to 24 % in the same period: −5 points with no intervention.
- Difference in differences: (−10) − (−5) = −5 points attributable to the protocol.
- Reading: half the improvement (5 of 10 points) was a general trend; the historical comparison doubled the effect.
What each type of comparison controls
| Design | Time trend | Differences between patients | Regression to the mean |
|---|---|---|---|
| Uncontrolled before-after | No | Yes (same group) | No |
| Historical controls | No | No | Partly |
| Concurrent non-random control | Yes | Not fully | Yes, if selected the same way |
| Randomised trial | Yes | Yes | Yes |
Markers
- Before-after
- Historical control
- Regression to the mean
When to use it
- Useful for generating hypotheses and for rare or lethal diseases with no earlier treatment.
- Common in quick evaluations of hospital protocols, new operations and devices.
- Before believing a before-after improvement, ask what else changed in that time.
Common pitfalls
- Selecting patients for an extreme baseline value (very high blood pressure) and crediting the treatment with the return towards the mean.
- Comparing with controls from another era with other diagnostic criteria or less sensitive tests.
- Concluding a treatment works without a concurrent control when the expected effect is moderate.
Test yourself
Think of the answer, then tap the question to see it.
Why do historical controls tend to exaggerate benefit?
Because general improvements in care, changes in criteria and less severe patients are credited to the intervention.
What is regression to the mean in a before-after study?
People selected for an extreme value tend to have a value closer to the average at the next measurement, even without treatment.
When can a historical control be acceptable?
When the effect is very large and the untreated course is well known and stable, such as a disease that was always fatal.
Connections
Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.
Sources: Chapter 24: Including non-randomized studies on intervention effects (Cochrane, 2024); Chapter 25: Assessing risk of bias in a non-randomized study (Cochrane, 2024); Difference-in-Difference Estimation (Columbia University Mailman School of Public Health, 2026).
Spotted a mistake?We are always improving for the community. Tell us what is wrong and our team will review it.Report a mistake
Representative model, for educational purposes only.