Letter to the Editor |
RE: "EASY SAS CALCULATIONS FOR RISK OR PREVALENCE RATIOS AND DIFFERENCES"
1 Clinical Epidemiology Research and Training Unit, Boston University School of Medicine, Boston, MA 02118
2 Section of Epidemiology and Preventive Medicine, Boston University School of Medicine, Boston, MA 02118
(e-mail: tneogi{at}bu.edu)
We read the Journal editorial by Drs. Spiegelman and Hertzmark (1
) with great interest and enthusiasm. Using log-binomial regression, one can directly estimate risk ratios and risk differences when disease occurrence is measured by using cumulative incidence (risk) in a cohort study. These effect estimates are easily communicated to and understood by the scientific community and the public. Furthermore, estimation of risk difference is of great importance in terms of public health issues.
Although the authors (1) did not advocate the use of prevalence ratios over prevalence odds ratios in cross-sectional studies, we hope that investigators will not abandon the prevalence odds ratio entirely. In fact, under certain circumstances, the prevalence odds ratio may be more appropriate than the prevalence ratio for assessment of exposure-disease relations. The prevalence odds of a disease, not the prevalence, is equivalent to the product of the incidence rate of the disease and the average duration of that disease in a cross-sectional study with a stationary population and no migration in and out of the prevalence pool (2). When the average duration of disease among the exposed group is the same as that among the nonexposed group, the prevalence odds ratio equals the incidence rate ratio. The prevalence ratio approximates the prevalence odds ratio only when the prevalence of disease is low (<10 percent). Therefore, only when the disease is rare can the prevalence ratio approximate the incidence rate ratio.
Thus, in the context of a cross-sectional study, if the assumptions of a stationary population, no migration, and equal average disease durations in the exposed and nonexposed are appropriate, we recommend the use of prevalence odds ratios obtained from logistic regression rather than prevalence ratios obtained from log-binomial regression.
ACKNOWLEDGMENTS
Conflict of interest: none declared.
References
- Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005;162:199–200.
[Free Full Text] - Greenland S, Rothman KJ. Measures of effect and measures of association. In: Rothman KJ, Greenland S, eds. Modern epidemiology. 2nd ed. Philadelphia, PA: Lippincott-Raven, 1998:47–64.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M R Petersen and J A Deddens A revised SAS macro for maximum likelihood estimation of prevalence ratios using the COPY method Occup. Environ. Med., September 1, 2009; 66(9): 639 - 639. [Full Text] [PDF]
|
|
|
|
|
|
J A Deddens and M R Petersen Approaches for estimating prevalence ratios Occup. Environ. Med., July 1, 2008; 65(7): 501 - 506. [Full Text] [PDF]
|
|
|
|
 |
|
 D. Spiegelman and E. Hertzmark THE AUTHORS REPLY Am. J. Epidemiol., June 15, 2006; 163(12): 1159 - 1161. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||