Over one in eight adults are now obese -- a ratio that has more than doubled since 1975 and will swell to one in five by 2025, a major survey reported Friday.

Of about five billion adults alive in 2014, 641 million were obese, the data showed -- and projected the number will balloon past 1.1 billion in just nine years.

The research warned of a looming crisis of "severe obesity" and disease brought on by high-fat, high-sugar diets causing blood pressure and cholesterol to rise.

"There will be health consequences of magnitudes that we do not know," author Majid Ezzati of Imperial College London told AFP.

The survey, published in The Lancet medical journal, claimed to be the most comprehensive of its kind conducted to date.

People are divided into healthy or unhealthy weight categories based on a universally-adopted measure dubbed Body Mass Index (BMI) -- a ratio of weight-to-height squared.

A healthy BMI ranges from 18.5 to 24.9.

One is considered underweight below 18.5, overweight from 25 up, and obese from 30 -- when the risk of diabetes, stroke, heart disease and some cancers escalates massively.

With a BMI of 35 one is categorised as severely obese, and from 40 upward as morbidly so.

Among men globally, obesity tripled from 3.2 percent of the population in 1975 to 10.8 percent in 2014 (some 266 million), and among women from 6.4 percent to 14.9 percent (375 million), said the survey -- 12.9 percent combined.

This was equivalent to the average adult, 18 and older, being 1.5 kilos (3.3 pounds) heavier every decade.

"If the rate of obesity continues at this pace, by 2025 roughly a fifth of men (18 percent) and women (21 percent) will be obese," according to a statement by The Lancet.

More than six percent of men and nine percent of women will be severely obese.

- Weighty flip-flop -

The ratio of underweight people in the world declined at a slower rate than obesity grew, said the authors -- from about 13.8 percent in 1975 to 8.8 percent for men, and 14.6 percent to 9.7 percent for women.

"Over the past 40 years, we have changed from a world in which underweight prevalence was more than double that of obesity, to one in which more people are obese than underweight," said Ezzati.

At current rates, more women will be severely obese (a BMI of 35 or more) than underweight by 2025, and the world will miss its stated target of halting obesity at 2010 levels.

In 2014, the world's fattest people lived in the island nations of Polynesia and Micronesia, where 38 percent of men and more than half of women were obese, said the study.

Nearly a fifth of the world's obese adults (118 million) lived in six high-income countries -- the United States, Britain, Ireland, Australia, Canada and New Zealand.

The US was home to one in four severely obese men and almost one in five severely obese women in the world.

- Surgery as a solution? -

At the other extreme, the paper said, Timor-Leste, Ethiopia and Eritrea had the lowest BMI numbers in the world, with averages as low as 20.1

More than a fifth of men in India, Bangladesh, Timor-Leste, Afghanistan, Eritrea and Ethiopia, and a quarter of women in Bangladesh and India, were underweight.

"The global focus on the obesity epidemic has largely overshadowed the persistence of underweight in some countries," the research paper said.

"To address this problem will require social and food policies that enhance food security in poor households, but also avoid overconsumption of processed carbohydrates and other unhealthy foods."

Like being underweight, severe and morbid obesity has many health risks.

"We can deal with some of these, like higher cholesterol or blood pressure, through medicines," said Ezzati.

"But for many others, including diabetes, we don't have effective treatment."

The paper says stomach-shrinking bariatric surgery may become the "most effective intervention for weight loss and disease prevention" as waistlines continue expanding.

The data was compiled from 1,698 studies involving 19.2 million adults from 186 countries which are home to 99 percent of the world's population.

Unlike earlier research, studies were only included if the participants' height and weight had been measured -- not self-reported.

The data did not include statistics on children and teenagers.