While being overweight increases one's chances of suffering a fatal medical problem, being underweight can be just as risky, says new research out of Toronto, Canada.

A study led by Joel Ray, a physician-researcher at St. Michael's Hospital and the hospital's Li Ka Shing Knowledge Institute, asserts there's a link between being underweight and a higher risk of dying, in both adults and fetuses, even when factors such as smoking, alcohol use or lung disease are considered, or adults with chronic or terminal illnesses are excluded.

Ray's new analysis included 51 studies on the links between body mass index and deaths from any cause, plus data on newborn weight and stillbirths in Ontario.

The re-examination of the data revealed adults who are underweight -- in other words, with a BMI under 18.5 -- have an estimated 1.8 times greater risk of dying than those with what Ray called a "normal" BMI of 18.5 to 24.9.

The risk of dying is 1.2 times greater, the research notes, for people who are obese, with a BMI of 30-34.9 or greater, and 1.3 times higher for those who are severely obese, with BMI of 35 or more.

"BMI reflects not only body fat, but also muscle mass. If we want to continue to use BMI in health care and public health initiatives, we must realize that a robust and healthy individual is someone who has a reasonable amount of body fat and also sufficient bone and muscle," Dr. Ray said in a news release. "If our focus is more on the ills of excess body fat, then we need to replace BMI with a proper measure, like waist circumference."

The researchers followed study subjects for five years or longer and attempted to separate those who died during that time because of cancer, chronic lung disease or heart failure.

Common causes of being underweight include malnourishment, heavy alcohol or drug use, smoking, low-income status, mental health or poor self-care.

Said Ray: "We have an obligation to ensure that we avoid creating an epidemic of underweight adults and fetuses who are otherwise at the correct weight. We are, therefore, obliged to use the right measurement tool."

The research has been published in the Journal of Epidemiology and Public Health.