|Home | About | Journals | Submit | Contact Us | Français|
Ezzati et al. (May 2006 JRSM1) misplace their confidence by a wide margin in claiming to have reached the first ‘unbiased’ estimates of state and national levels and trends of obesity in the USA. To do so, they claim to have calculated rates of self-reporting biases using data from the NHANES (National Health and Nutrition Examination Survey) and Behavioral Risk Factor Surveillance System (BRFSS) surveys of the US Center for Disease Control and Prevention, which were assumed to be representative of the USA population on the basis of the original sampling design. In reality the assumption is questionable, the response rates being about 80% for NHANES and 45% for BRFSS. If nothing else, such a discrepancy sets a distance between what the two surveys may represent, also considering that the surveys represent different samples. In a crucial decision, Ezzati et al. excluded subjects who did not participate in both surveys—an exclusion that inevitably invalidates any claim that the bias estimates reached could be representative of the USA population or of individual state populations. Such elementary flaws lead to ask how scarce resources could be approved for such a wanting study design.
Competing interests None declared.