Copyright © 2005 by the Johns Hopkins Bloomberg School of Public Health
Letters to the Editor |
THE AUTHORS REPLY
1 Institute of Medical Epidemiology, Biometry and Epidemiology, University Hospital of Halle, 06097 Halle (Saale), Germany
2 Institute of Medical Informatics, Biometry and Epidemiology, University Hospital of Essen, 45147 Essen, Germany
We congratulate Dr. Voigt et al. (1), who, triggered by our recent publication (2), performed analyses of exposure prevalence and exposure misclassification by recruitment wave in a subvalidation analysis of their case-control study of breast cancer through linkage of interview data with pharmacy records from a Washington State health maintenance organization (3).
Voigt et al. conclude that the "sensitivity and specificity of self-reported data on antihypertensive drug use were slightly lower for cases interviewed 2 or more months after first contact than for those interviewed promptly" (1, p. 401). We do not agree with the word "slightly" here. Although the number of subjects in the subvalidation study was quite low and point estimates of sensitivity and specificity were imprecise, the point estimates (the best estimates derived from the data) show that sensitivity and specificity decreased by 8–10 percentage points from early-responding cases to late-responding cases and by only 1–4 percentage points among controls. These data indicate that the exposure misclassification increases differentially by wave by an amount which may have effects on the odds ratio estimates. In other words, early-responding cases and controls show similar sensitivities and specificities. However, late-responding cases have a lower sensitivity and specificity than do controls, which represents, by definition, differential exposure misclassification.
In our original paper (2), we did not present analyses dealing with the effect of differential exposure misclassification error by wave on point estimates in cohort or case-control studies. Obviously, Voigt et al. (1) raised a more complicated point that can be illustrated by recalculating "observed" cell counts from the true cell counts reported by Voigt et al. (see table ). It is intriguing to see that the exposure misclassification bias of the odds ratio points away from the null among early respondents and points toward the null among late respondents. Therefore, no clear interpretation of the results is at hand, notwithstanding the problem of small numbers.
|
The reader should be warned not to simply apply our results from cohort studies to case-control studies. At the moment, the calculus of wave-specific misclassification and response in the case-control setting is not completely understood and needs further investigation.
 |
REFERENCES |
|---|
 TOP REFERENCES |
|---|
- Voigt LF, Boudreau DM, Weiss NS, et al. Re: "Studies with low response proportions may be less biased than studies with high response proportions." (Letter). Am J Epidemiol 2005;161:401–402.
[Free Full Text] - Stang A, Jöckel KH. Studies with low response proportions may be less biased than studies with high response proportions. Am J Epidemiol 2004;159:204–10.
[Abstract/Free Full Text] - Boudreau DM, Daling JR, Malone KE, et al. A validation study of patient interview data and pharmacy records for antihypertensive, statin, and antidepressant medication use among older women. Am J Epidemiol 2004;159:308–17.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. Olson Survey Participation, Nonresponse Bias, Measurement Error Bias, and Total Bias Public Opin Q, January 1, 2006; 70(5): 737 - 758. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||