Vince and Larry are the NHTSA’s famous crash test dummies. They are both “50th percentile adult males.” That means they would stand 5’9” tall and weigh 171 pounds. But researchers at the University of Virginia point out that designing cars for Vince and Larry may leave those of us shaped differently at risk during a crash. Half of all licensed drivers are female and—according to the CDC—40% of Americans are obese. So UVA came up with an all-new safety study.

UVA studied both drivers in 1,700 crashes (so 3,400 drivers total) that occurred between 1996 and 2008. It chose to study pairs involved in the same crash to control for factors such as severity of the crash and emergency services’ response times. It also narrowed the study to crashes between similarly-sized vehicles. The results are fascinating—and dramatic.

For example, women wearing seat belts are 47% more likely to suffer severe injury than men wearing seat belts. So perhaps we need to revisit our seat belt design to better protect female drivers.

One of the highest disparities in the study was the risk of death for morbidly obese drivers—those with a BMI over 40. They were 80% more likely to die in a crash than drivers who weren’t obese.

Researchers analyze the increased fatality risk for obese drivers

Professor Jason Kerrigan was the deputy director of UVA’s Center for Applied Biomechanics when it completed the study. He admits he had a hypothesis before the study—but was proven wrong. “One of our assumptions was that in a frontal crash, obese people are more likely to submarine under the lap belt, causing a higher incidence of abdominal injurie[s]… While we found this not to be the case, we have noticed an increase in lower extremity injuries.”

Scientific American pointed out that seat belts held back obese and non-obese drivers’ upper bodies equally effectively. But the morbidly obese drivers’ pelvises moved forward more before their lap belt stopped them. Still, researchers are unsure whether this is the cause of the increased fatality rate—or pre-existing conditions that often co-occur with morbid obesity. Kerrigan admitted, “We don’t understand enough about the nature of obesity to know why it makes the situation worse.”

Kerrigan concluded that more data is needed—and with new methods, such as UVA’s study. “For years, we used a technique called geometric scaling to forecast how human beings of different sizes would respond to crashes.” But that system simply, “doesn’t work.”