Thursday, April 21, 2005

Estimates of Obesity-Related Deaths Falling Rapidly; Smoking Remains Overwhelmingly the Leading Cause of Preventable Death

A paper published in Wednesday's issue of JAMA (Journal of the American Medical Association) reported that obesity causes an estimated 112,000 deaths per year in the United States, down markedly from the estimate of 400,000 reported in an article in the same journal in March 2004.

Authors from the Centers for Disease Control and Prevention (CDC) originally reported last March that overweight causes an estimated 400,000 annual deaths, although this figure was later reduced to 365,000 due to a computational error. Now, authors from CDC, UC Berkeley, and the National Cancer Institute report, using a different methodology and more recent data, that the estimated deaths attributable to obesity are 112,000 per year. However, being overweight (but not obese) was associated with 86,000 fewer deaths; thus, the paper reported an estimate of 26,000 annual deaths attributable to excessive weight (overweight and obesity combined). The precision of this estimate was low enough, however, that a conclusion of 0 excess deaths from excessive weight could not be ruled out.

The Rest of the Story

In reporting their estimate of 400,000 annual overweight-related deaths, the authors of the original JAMA findings claimed that "poor diet and physical inactivity may soon overtake tobacco as the leading cause of death." Not surprisingly, the media as well as medical and public health organizations widely disseminated this conclusion as being the main finding of the research. Typical headlines read: "Obesity to Overtake Smoking as Leading Cause of Death," "CDC Reports Obesity Overtaking Smoking as No. 1 Killer," "Obesity Now Rivals Smoking as Killer," and "Poor Diet, Inactivity Becoming a Leading Cause of Death."

Ten months following publication of these findings, the authors reduced the estimate to 365,000 deaths a year due to a computational error. Now, a more rigorous methodology has produced an estimate of 112,000 excess annual deaths due to obesity, but only 26,000 excess deaths due to excessive weight. Due to imprecision in the calculation, the authors could not conclude that this figure was statistically different from 0. Thus, after all the fanfare surrounding the original report and its prediction that overweight would soon overtake smoking as the leading cause of death, it is now not even clear that the adjusted overall death rate for individuals with excessive weight exceeds that for normal weight individuals.

There are a number of improvements in the methodology for estimation of obesity- and overweight-attributable deaths in the more recent JAMA paper that suggest Wednesday's estimate is more reliable than the March 2004 figure (as opposed to these estimates simply reflecting chance variation). First, and most importantly, the 2005 paper incorporates more recent mortality data: all three NHANES (National Health and Nutrition Examination Survey) cohorts were used, which means that follow-up data were available through 2000 rather than just through 1992. Mortality risk associated with obesity and overweight was markedly higher in NHANES-I than in the more recent NHANES-II and NHANES-III.

Second, the 2005 paper accounts more fully for the fact that the mortality risk associated with obesity decreases with age. If not properly accounted for, this will result in inflated estimates of deaths because mortality risk estimates derived from younger individuals cannot be accurately applied to older individuals whose actual obesity-related risks are considerably lower.

What is clear from Wednesday's article is that obesity is a major public health problem that causes substantial mortality. However, more modest levels of excess weight do not appear to be a major health problem; in fact, they may produce lower mortality risk. What this likely means is that national standards for defining overweight may have to be changed. They likely need to be increased to reflect the apparent lack of current evidence for harmful sequelae of what is now considered to be modest levels of excess weight.

What is also clear is that neither excessive weight, however it is defined, nor obesity is even close to causing the tremendous death toll attributable to tobacco products.

Finally, this story underscores the importance of public health researchers coming to terms with the limitations of the methods available to them, and taking great steps not to over-interpret study results when these methods are substantially limited. This is especially important for quantitative estimates. While the only real research danger of over-interpretation is quantitative inaccuracy, there is a more important danger of producing inaccurate media and other reports that may inappropriately influence public policy, such as, in this case, the allocation of scarce resources.


1. Mokdad AH, Marks JS, Stroup DJ, Gerberding JL. Actual causes of death in the United States, 2000. JAMA 2004;291:1238-1245.
Mokdad AH, Marks JS, Stroup DJ, Gerberding JL. Correction: Actual causes of death in the United States, 2000 (letter). JAMA 2005;293:293-294.
3. Flegal KM, Graubard BI, Williamson DF, Gail MH. Excess deaths assocaited with underweight, overweight, and obesity. JAMA 2005;293:1861-1867.

UPDATE (April 28, 2005; 2:20 pm): In an excellent article on this topic on the American Council for Science and Health's (ACSH) Facts and Fears site, Dr. Ruth Kava, Director of Nutrition at ACSH points out that my conclusion that overweight may not be a public health problem, based on the reduced mortality risk in this group, may be inaccurate and unwise for two reasons. First, some individuals who are in excellent physical shape because they exercise regularly may actually have increased body muscle mass which places them into the higher BMI group, thus making it appear that overweight is protective (note that BMI does not distinguish between lean body tissue and fat). Second, looking only at mortality ignores the other potentially harmful sequelae of overweight in terms of morbidity - including diabetes, hypertension, and osteoarthritis. Please see Dr. Kava's post for a nice discussion of this issue.

See also Associate Director of ACSH Jeff Stier's excellent article on the obesity mortality estimates and CDC's handling of this situation.

No comments: