Figure 2.

Genetic prediction of body weight—context matters

A 49-year old woman, who eats a balanced diet, runs 6 miles/day, and commutes by bike to work, does a direct-to-consumer genetic test with one of the many online personalized genomics companies. The company claims to provide genetic insight into her health to make it easier for her to take action. She provides a saliva sample and completes numerous questionnaires on her physical and mental health, family medical history, and more. Her reported results show that, based on the genetic variants tested, she carries 376 weight-lowering variants and 332 weight-increasing variants, predisposing her to weigh about “average” or 157 lbs (71.2 kg, based on the company’s customers’ weight of the same age, height, and sex). However, the woman’s real weight was 120 lbs (54.4 kg). A likely reason for why this genetic test overpredicted the woman’s weight by 30% is because her lifestyle—even though information was shared in detail—was not appropriately incorporated in the prediction models.