By: Katherine Crawford


The ultimate goal of scientific research is to generate new knowledge that can be applied to benefit as many people as possible.  But this goal may be challenging to achieve based on who is included in research studies. 


Imagine what could happen if medical interventions were only tested on certain groups of people?  These interventions could be ineffective or down right harmful when applied to other populations. This is why researchers strive to have representative samples – where all members of a population are accounted for in a smaller study sample.  By having representative samples, we can better understand if certain individuals or populations react differently to therapies and interventions.

Years ago, the NIH and others in the scientific community noted that medical research was often not representative of women or people of color. As a result, the NIH took steps to fund research into underserved populations.

But might there be other populations who are still not well represented in scientific research?  And what do we know generally about who participates in scientific research?

People who participate in ongoing research studies may be unique.

The Avon Longitudinal study of Parents and Children (ALSPAC) is a large cohort study from the UK that we often use in our own research in the Dunn Lab.  ALSPAC began their study back in the 1990’s by capturing 75% of all pregnant mothers living in a particular county. Over the next 20 years, ALSPAC researchers followed these moms and their children, asking them questions about their life, taking biological samples, and measuring their overall health and wellbeing. As expected, many people stopped responding to the study over these 20 years.  In fact, by late childhood, only about 53% of the original moms were returning mailed questionnaires.

So, what happened to those who left the study? Is there anything different about the 47% of the moms who stopped responding to the study compared to the moms who kept participating?

ALSPAC investigators learned that the mothers who left the study were not a random assortment of the original group. In fact, ALSPAC researchers could predict who returned survey questionnaires based on socioeconomic questions asked at the beginning of the study. Things like marital status, family size, and maternal education level all predicted whether surveys were returned – even up to 10 years later.

There has even been evidence that parents of children displaying traits associated with things like opposition defiant disorder were less likely to return surveys or attend clinics as time went on.  The patterning of drop-out by these social factors is a problem because it can lead to biases in terms of who researchers are able to study and generalize their study findings to.

But genetic research is fine, right?

Unfortunately, no. Another thing ALSPAC investigators learned is that there are genetic factors that also influence whether or not someone will return a survey.

ALSPAC compared the genetic makeup of mothers and children who stayed in the study to those who dropped out and found some interesting results. For example, children who had a genetic predisposition to depression (based on a “polygenic risk score”, meaning a genetic measurement that uses variation across multiple genetic variants to predict a health outcome) were more likely to drop out of the ALSPAC study. There was also a particularly strong correlation between genetic predisposition to schizophrenia and study drop-out even after accounting for the effects of the social factors mentioned earlier.

So, what does this mean for future research?

Because there are factors that influence who will participate and remain in scientific studies, researchers may end-up excluding the specific groups that are already predisposed to be less likely to participate.

For example, people with psychiatric disorders may be underrepresented in research studies, especially longitudinal studies or studies with multiple visits. Conceptually this makes sense since people experiencing depression, for instance, may find it hard to drive to the clinic for a research study or answer a long, extensive questionnaire.

Anything that makes it harder for a person to participate in a research study – whether that is health, time, money, or lack of child care– will make it more likely that they will stop participating.  As a result, valuable information can be lost about the experiences of people who have barriers to study participation.

As scientists, we must be aware of the effects of attrition in our research, both at the start of our studies and as they progress.  Of course, there are innovative methods that can be used to reduce the ways in which study drop-out can create biased estimates of the associations we are looking for.  But perhaps it is better to start by designing studies that make it easier for people to participate so that no one slips through the cracks and science can be used for the benefit of everyone.