The report, by far the largest into BMI and mortality, analysed nearly 100 studies.

Experts say the report indicates that overweight people need not panic unless they have other indicators of poor health, and that depending on where fat is in the body, it might be protective or even nutritional for older or sicker people.

Dr Steven Heymsfield of the Pennington Biomedical Research Centre, in Louisiana, who wrote an editorial accompanying the study, said that for overweight people, if such indicators as cholesterol were ''in the abnormal range'', then that weight was affecting you. But if indicators were normal, there was no reason to go on a crash diet.

But piling on kilograms and becoming more than slightly obese did remain dangerous.

''We wouldn't want people to think, 'Well, I can take a pass and gain more weight','' said Dr George Blackburn, associate director of Harvard Medical School's nutrition division.

Rather, he and others said, the report, in The Journal of the American Medical Association, suggested BMI, a ratio of height to weight, should not be the only indicator of healthy weight.

Dr Samuel Klein, the director of the Centre for Human Nutrition at Washington University School of Medicine in St Louis, said: ''Body mass index is an imperfect measure of the risk of mortality.''

Such factors as blood pressure, cholesterol and blood sugar must be considered, he said. Experts also said the data suggested the definition of ''normal'' BMI, 18.5 to 24.9, should be revised, excluding its lowest weights, which might be too thin.

While obese people had a greater mortality risk overall, those at the lowest obesity level (BMI of 30 to 34.9) were not more likely to die than people of normal weight.

But ''Once you have higher obesity, the fat's in the fire'', Dr Blackburn said.

NEW YORK TIMES