Researchers build a statistical model using family health history to improve disease risk assessment

Published in March in the peer-reviewed journal BMC Medical Research Methodology, the results may advance physicians’ ability to use family health history information to assess which conditions with a genetic component pose a risk to your health and require further testing.

The researchers used type 2 diabetes mellitus (T2D) as the disease model for their study. An estimated 30 million people in the US suffer from the condition, making it one of the most common and costly chronic diseases to treat. T2D is known as a complex disease because its development depends strongly on a person’s genetic predisposition, combined with their environment and lifestyle.

“Family health history is one of the most powerful predictors of the risk a person carries for forming type 2 diabetes,” says Dr. Chris Marcum, NHGRI staff scientist and senior author of the paper.

Doctors typically receive family health history from their patients, who may not have complete or accurate knowledge of their relative’s diagnoses, age at diagnosis, lifestyle and cause of death. Reports from patients about their T2D family history range widely from 53 to 87% accuracy. Patients who inaccurately discuss their family health history of T2D may cause doctors to either over- or under-predict the patient’s risk for the disease. Such concerns are especially true for younger members of a family who have yet to learn about their family history.

Family Health History animation. Credit: Ernesto Del Aguila III, NHGRI.

Current research has attempted to resolve this issue, but with several constraints. Most studies on improving family health history information are piecemeal and specific, relying on models that can only be used for complex diseases such as hypertension, high cholesterol and heart disease. In contrast, Dr. Marcum and his colleagues aimed to create a statistical model that could be used as a general tool for assessing disease risk for all types of complex diseases.  

The researchers collected family health history from 45 families (with multiple participants per family) residing in the Cincinnati area. Family members were asked to report their own history of T2D and for their immediate relatives such as parents, siblings, children and grandparents. They also provided health behavior information such as alcohol and smoking habits and weight.

“Health behaviors like smoking and alcohol use can increase risk for type 2 diabetes, while exercise and a healthy diet can reduce that risk, so it was important to include that information in our model,” says Dr. Marcum.

The researchers used a statistical model that integrated family members' reports of who did and who did not have T2D (called T2D status). The model accounted for the fact that individuals from the same family may have similar reports and similar family environments, based on common lifestyle and health behaviors. All of this information was pooled together by the model to make predictions about T2D status for all family members, and assess individual risk for unaffected members. The model had a 78% accurate prediction rate, a vast improvement from the 58-87% range from single patient family history reports.

But the current model has its limitations. Researchers need to replicate findings in other disorders. Since many families in the study had consistent knowledge about T2D status for family members, the researchers will need to evaluate how well the model stands up to highly inconsistent family health history reports.

When asked about the possible implementation of these results, Dr. Marcum responded, “There are privacy concerns that are embodied in current laws regarding the sharing of family health history that make implementation in a clinical setting a challenge. It will be necessary to have a meaningful discussion on how we can integrate these models into real-life clinical settings.”

In the near future, Dr. Marcum believes it will be necessary to create health record systems that are family-centric and also work with the medical privacy provisions set up by the Health Insurance Portability and Accountability Act of 1996 (HIPAA). “These limitations are opportunities for future research,” says Dr. Marcum.