American Journal of Epidemiology Advance Access originally published online on November 9, 2005
American Journal of Epidemiology 2006 163(1):99; doi:10.1093/aje/kwj004
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Letter to the Editor |
RE: "BODY MASS INDEX AND INCIDENT ISCHEMIC HEART DISEASE IN SOUTH KOREAN MEN"
Department of Health and Nutrition Sciences, Brooklyn College of the City University of New York, Brooklyn, NY 11210
(e-mail: jamesg{at}brooklyn.cuny.edu)
In a recent Journal article, Jee et al. (1
) may have found higher hazard ratios for a body mass index (BMI) of 20–30 kg/m2 than those found in Caucasian populations (e.g., Hu et al. (2)) primarily because the authors corrected two statistical biases more effectively than is usually done in similar studies with Caucasian cohorts.
The first bias is reverse causation, which artificially increases mortality risk at low BMI levels (3) and hence artificially deflates relative morality risk for a BMI of 20–30 kg/m2. Reverse causation is thought to be caused by factors that induce weight loss and simultaneously increase mortality risk, such as serious illness (3). Jee et al.'s (1) analysis used two techniques that correct for reverse causation and tend to increase the hazard ratios for a BMI of 20–30 kg/m2. They excluded participants with serious illness, and they excluded participants who had experienced weight loss—those whose BMI decreased more than two units in the 2 years prior to their follow-up baseline. Most survival analyses involving Caucasian cohorts have used only the first technique.
The second bias is the regression-dilution bias, which is due to regression-to-the-mean patterns in BMI data (4). These patterns occur because participants with a low BMI reported in a survey tend to have a higher usual BMI, and vice versa. Hence, if the unbiased relation between mortality risk and usual BMI is monotonically positive, mortality risk in a survival analysis will be artificially increased for low baseline BMI and will be decreased for high baseline BMI. Jee et al. (1) corrected for this bias by using the average of the 1990 and 1992 BMI measurements as their predictor, which would tend to increase the hazard ratios for a BMI of 20–30 kg/m2 (5). Very few survival analyses involving Caucasian cohorts have corrected for this bias.
It seems likely, therefore, that Jee et al.'s findings (1) do not provide reliable evidence that the hazard ratio–BMI relation is different in South Korean than in Caucasian populations.
ACKNOWLEDGMENTS
Conflict of interest: none declared.
NOTES
Editor's note: In accordance with Journal policy, Jee et al. were asked whether they wanted to respond to this letter, but they chose not to do so.
References
- Jee SH, Pastor-Barriuso R, Appel LJ, et al. Body mass index and incident ischemic heart disease in South Korean men and women. Am J Epidemiol 2005;162:42–8.
[Abstract/Free Full Text] - Hu FB, Willett WC, Li T, et al. Adiposity as compared with physical activity in predicting mortality among women. N Engl J Med 2004;352:2694–703.
- Manson JE, Stampfer MJ, Hennekens CH, et al. Body weight and longevity: a reassessment. JAMA 1987;257:353–8.
[Abstract/Free Full Text] - Clarke R, Shipley M, Lewington S, et al. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol 1999;150:341–53.
[Abstract/Free Full Text] - Greenberg JA. Removing confounders from the relationship between mortality risk and systolic blood pressure at low and moderately increased systolic blood pressure. J Hypertens 2003;21:49–56.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||