Exposure misclassification

Other factors influence the magnitude of the effect of exposure misclassification on estimates of association between exposures and disease. The effect depends not only on the extent of exposure misclassification, but also on the prevalence of exposure in the population studied. Since pesticide exposure prevalence may differ in different populations and is certainly different in general population stndies when compared to studies in farming communities, the performance of exposnre assessment techniques will vary according to the study context. The specificity determines the bias in risk-ratio situations with a low exposure prevalence. Thus, a poor sensitivity, for instance, the one reported by Arbuckle et al. (2002) for 2,4-dichlorophenoxyacetic acid (2,4-D), may not be problematic in a general population or case-control study, as long as the specificity is sufficiently high. 

Exposure misclassification may not be random, but may differ for individuals with and without the disease of interest (differential exposure misclassification). This further complicates the impact of exposure misclassification and may make it impossible to predict the effects without more detailed information about the structure of the errors. 

In an 11-year observational study in new users of non-selective, non-aspirin NSAIDs (n — 181 441) and an equal number of non-users there was no evidence of a protective effect of naproxen (50). During 532 634 person-years of follow-up there were 6382 cases of serious coronary heart disease (11.9 per 1000 person-years). Multivariate-adjusted rate ratios for current and former use of non-aspirin NSAIDs were 1.05 (95% Cl = 0.97, 1.14) and 1.02 (0.97, 1.08) respectively. Rate ratios for ibuprofen, naproxen, and other NSAIDs were 1.15 (1.02, 1.28), 0.95 (0.82, 1.09), and 1.03 (0.92, 1.16) respectively. There was no protection in long-term users with uninterrupted use the rate ratio among current users with more than 60 days of continuous use was 1.05 (0.91,1.21). When naproxen was directly compared with ibuprofen, the rate ratio in current users was 0.83 (0.69, 0.98). This study therefore seems to have shown no cardioprotective effect of naproxen. However, the study had a number of important limitations, including lack of information about some important confounders (smoking, obesity), possible exposure misclassification, and lack of information about over-the-counter use of aspirin. 

Comparison of the ratings of experienced raters with previously recorded industrial hygiene measurements for occupations in Australia Estimation of the levels of exposure misclassification by expert assessment in a study of lung cancer in central and eastern Europe and Liverpool Application of Bayesian framework for retrospective exposure assessment of workers in a nickel smelter Determination of the level of information required by industrial hygienists to develop reliable exposure estimates Explanation of new framework to obtain exposure estimates through the Bayesiem approach Validation of a new method for structured subjective assessment of past concentration... 

Blair, A., Stewart, R, Lubin, J. H., and Forastiere, F. (2007). Methodological issues regarding confounding and exposure misclassification in epidemiological studies of occupational exposures. Am J Ind Med 50, 199-207. 

Grandjean, P., Budtz-Jorgensen, E., Keiding, N., and Weihe, P. (2004). Underestimation of risk due to exposure misclassification. Int J Occup Med Environ Health 17, 131-136. 

Gustafson, P., and Greenland, S. (2006). Curious phenomena in Bayesian adjustment for exposure misclassification. Stat Med 25, 87-103. 

L., Krizanova, D., Cassidy, A., van Tongeren, M., and Boffetta, P. (2003). Assessing exposure misclassification by expert assessment in multicenter occupational studies. Epidemiology 14, 585-592. 

Nieuwenhuijsen, M. J., Grey, C. N., and Golding, J. (2005). Exposure misclassification of household pesticides and risk perception and behaviour. Ann Occup Hyg 49,103-109. 

The uncertainties and limitations in exposure assessment in these studies can result in exposure misclassification, which will lessen the ability to detect significant dose-response associations and might result in inaccuracies in the derivation of dose-response relationships. 

If exposure misclassification occurred in the studies of MeHg, such misclassification would tend to obscure any trae effect. Therefore, statistically significant dose-response associations are likely to reflect tme dose-response relationships, assuming that other sources of bias are adequately addressed. 

In epidemiology, ongoing efforts to deal with exposure misclassification should be continued. One particularly useful area of exploration might be to seek to understand the extent to which using smaller geographic areas (in terms of total size, and in terms of population density) instead of cities, in multi-locality epidemiology studies, may cause less exposure misclassification. 

---