CDC Letter Cites Computer Error for Problems With Study on Annual Obesity-Related Deaths
A computer error caused CDC in a March 2004 study to overstate by about 35,000 the number of annual obesity-related deaths in the United States, according to a letter from four agency researchers published on Wednesday in the Journal of the American Medical Association, the AP/Boston Globe reports. The study, published in JAMA, found that annual obesity-related deaths increased by 100,000 between 1990 and 2000 to 400,000 (Yee, AP/Boston Globe, 1/19).
For the study, Ali Mokdad, chief of the behavioral surveillance branch at CDC, and colleagues reviewed 2000 U.S. mortality data and studies on the role that lifestyle factors have on the development of conditions such as diabetes or stroke to estimate the number of deaths attributable to lifestyle. Researchers also compared their data with a similar study conducted with 1990 U.S. mortality data.
The study found that an estimated 400,000 U.S. residents died from causes related to dietary habits and sedentary lifestyles -- 33% more than in 1990, when obesity contributed 300,000 deaths. In comparison, the study found that 435,000 U.S. residents died from smoking or exposure to tobacco in 2000, compared with 400,000 in 1990. Researchers concluded that obesity would become the leading cause of preventable death by this year. The study led to an HHS campaign against obesity and an increased focus on obesity research at NIH (California Healthline, 11/23/04).
According to the letter, however, obesity will not overtake tobacco as the leading cause of preventable death by this year. CDC director Julie Gerberding, one of the four authors of the study, said a software malfunction contributed to computational errors that prevented the update of spreadsheets with data from earlier years with new data from 2000. CDC officials said that the agency plans to implement a new research agenda to find the most effective method to determine causes of death in large studies.
Gerberding, in response to criticism of the study in spring 2004, called for an internal agency investigation. CDC has completed the investigation but has not released a final report. CDC officials have said that they will post a summary of the report on the agency Web site "soon."
Sources familiar with the situation said the report will recommend that CDC implement an electronic system to improve access to research and work to develop an improved definition for obesity-related deaths (McKay, Wall Street Journal, 1/19). Some experts estimate that CDC overestimated the number of annual obesity-related deaths by more than 100,000 in the March 2004 study (Mestel, Los Angeles Times, 1/19).
Gerberding in a statement said that CDC will implement a "broader research agenda" with government, academic and industry researchers to determine how to measure the health effects of nutrition and physical activity. CDC officials have said that concerns from agency researchers who questioned the March 2004 study before publication "weren't heeded," the Journal reports. "Integrity is a core value of CDC, and the integrity of our science must be protected," Gerberding said (Wall Street Journal, 1/19).
Mokdad said that regardless of the problems with the study, the number of annual obesity-related deaths in the United States has increased in recent years (Los Angeles Times, 1/19).
Rep. Henry Waxman (D-Calif.), who has requested a Government Accountability Office investigation of the study, said, "CDC needs to disclose how a flawed paper was cleared within the agency and what measures are in place to prevent similar problems in the future. I am pleased that the agency has pledged to pursue better methods to assess the harm of obesity, a serious public health problem. This effort needs to move quickly" (Wall Street Journal, 1/19).
University of California-San Francisco medical professor Stanton Glantz said that CDC officials are "still stonewalling and still denying the fundamental problem" with the study (Los Angeles Times, 1/19).
An excerpt of the CDC letter published in JAMA is available online.