American Journal of Epidemiology Advance Access originally published online on November 17, 2007
American Journal of Epidemiology 2008 167(1):123-124; doi:10.1093/aje/kwm347
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
LETTERS TO THE EDITOR |
THE AUTHORS REPLY
1 Department of Health Policy and Management, Center for Injury Research and Policy, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD 21205
2 Yale Occupational and Environmental Medicine Program, Departments of Internal Medicine and Epidemiology and Public Health, Yale University School of Medicine, New Haven, CT 06510
(e-mail: kpollack{at}jhsph.edu)
We thank Dr. Brooks for her comments (1). Dr. Brooks raises an issue related to the age x body mass index (BMI) interaction term included in the multivariate logistic regression model presented in table 3 of our paper (2). She points out that we listed the multivariate-adjusted odds ratios for the main-effect BMI terms without providing information about the effect size of the age x BMI interaction term. Dr. Brooks is correct in stating that we did not include comments on the influence of the interaction term when we interpreted our results. However, in the Materials and Methods section of our paper (2), we reported that statistical significance testing for the interaction term yielded a p value of 0.04. Since the p value for the interaction term was less than the alpha level of 0.05 specified a priori, it was considered statistically significant and was included in the final multivariate model. Despite the fact that the interaction term was statistically significant, when we ran the logistic regression analyses in models stratified by age, there was no appreciable difference between the point estimates for the various age categories. Therefore, we decided not to present the odds ratio for each age category, since the odds ratios were essentially all the same.
It is likely that if the age x BMI interaction term had been highly significant, instead of marginally significant, the odds ratios would indeed have varied across the age categories. Readers should also note that if we had chosen an alpha level of 0.01 for statistical significance in this analysis, the age x BMI interaction term would not have been statistically significant.
While Dr. Brooks is correct in implying that more research needs to be conducted to truly understand the impact of BMI on traumatic workplace injuries, corrective action related to the modeling in this paper is not the necessary step. Instead, what is needed is more research exploring the various mechanisms of injury in hopes of developing appropriate strategies for prevention (3).
 |
ACKNOWLEDGMENTS |
|---|
Conflict of interest: none declared.
|
References |
|---|
 TOP References |
|---|
- Brooks MM. Re: "Association between body mass index and acute traumatic workplace injury in hourly manufacturing employees." (Letter). Am J Epidemiol (2008) 167:123.
[Free Full Text] - Pollack KM, Sorock GS, Slade MD, et al. Association between body mass index and acute traumatic workplace injury in hourly manufacturing employees. Am J Epidemiol (2007) 166:204–11.
[Abstract/Free Full Text] - Pollack KM, Cheskin L. Obesity and workplace traumatic injury: does the science support the link? Inj Prev (2007) 13:297–302.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||