In recent news, you may have heard of different terms around bias in different studies related to Huntington’s Disease. We wanted to take the time to walk through these definitions and examples of how those biases may affect a clinical trial.
Please note this is not based on any current or past clinical trial data but is meant to help educate community members on these terms in general.
When a Huntington’s disease (HD) clinical trial uses data from a natural-history study as an external control instead of enrolling a concurrent placebo group, researchers must consider how differences between the two groups could affect the interpretation of the results.
One important consideration is attrition bias, including a related concern sometimes described as survivorship bias.
What is attrition bias?
Attrition bias can occur when people who leave a study, miss assessments, or otherwise have incomplete follow-up differ in important ways from those who remain.
For example, imagine an HD clinical trial with:
- Treatment group: 50 people receiving an investigational therapy
- External control group: 50 people whose data come from a natural-history study
Over four years, some participants in the natural-history study may no longer have complete assessments. They may have:
- stopped participating,
- been unable to complete assessments because of disease progression,
- missed study visits,
- moved or changed care,
- or otherwise had incomplete follow-up.
The key question is: Are the people who remain in the dataset representative of the original group?
For example, suppose people whose HD is progressing faster are more likely to stop participating in a natural history study. The people who remain could be, on average, those progressing more slowly. The control group would look like it’s declining more slowly than the full group really was. As a result, a treatment would look less effective that it truly is, because it’s being compared against a control group that appears healthier than it should.
This is one reason missing data are an important consideration when interpreting externally controlled trials. The assumptions researchers make about why data are missing can influence estimates of treatment effects.
Where does survivorship bias come in?
Survivorship bias is a related concept. It occurs when conclusions are based disproportionately on people who are available for observation, while information about people who drop out or were never able to participate in the first place are not included.
In an HD natural-history study, this could matter if participants who can engage in a natural history study may fundamentally be different that those that cannot.
This does not mean that natural-history studies are inherently unreliable. Rather, it means that researchers need to understand who is represented in the available data and who is missing, and account for those differences when designing and analyzing an external-control comparison.
Why is a concurrent placebo group different?
A concurrent placebo group is part of the same randomized clinical trial as the treatment group. Participants generally:
- meet the same eligibility criteria,
- are assessed on the same schedule,
- use the same outcome measures,
- are followed under the same protocol, and
- undergo the same trial procedures.
Most importantly, randomization helps distribute known and unknown factors that could influence disease progression between the treatment and placebo groups.
An external control does not receive that benefit of randomization. Even when researchers carefully match participants or use statistical methods to adjust for differences, there may be characteristics that are difficult or impossible to measure and account for completely.
What if a placebo-controlled trial presents ethical concerns?
The question becomes more complicated when a conventional placebo-controlled trial may raise ethical concerns.
For some investigational therapies—particularly therapies involving invasive procedures, surgery, or other significant risks—researchers and regulators may need to consider whether a traditional placebo or imitation procedure is appropriate.
In those circumstances, an external control based on natural-history data may be a reasonable alternative. But using an external control does not eliminate the need to carefully evaluate differences between the populations, missing data, follow-up, assessment methods, and other factors that could influence the comparison.
Why this matters in HD research
For a disease such as Huntington’s Disease, where progression can vary considerably between individuals and clinical trials may follow participants for several years, the quality and completeness of long-term data are particularly important.
An external-control approach can provide valuable information, particularly when a concurrent placebo group is difficult or potentially problematic to use. At the same time, researchers, regulators, and the HD community need to understand the limitations of comparing participants from different studies.
The central question is not simply “Was there a placebo group?”
It is also:
“How comparable are the people in the treatment and external-control groups, and how might differences in follow-up or missing data affect the results?”
Understanding these issues can help the HD community better interpret clinical-trial results and the evidence used to support regulatory decisions.