Unbiased vs. conservative estimators of etiological fractions: examples of misclassification from studies of occupational lung cancer

Our plain-English summary of Unbiased vs. conservative estimators of etiological fractions: examples of misclassification from studies of occupational lung cancer, by O. Olsen (American Journal of Industrial Medicine, 1995).

The study

• Author: O. Olsen
• Journal: American Journal of Industrial Medicine
• Published: 1995;27(6):837-43
• Cited by: 4 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

Theoretical studies emphasize the importance of making unbiased etiological fraction estimates. In empirical works, however, the published estimates are usually conservative. The purpose of the present report is to study, empirically, the numerical magnitude of such conservative biases. Examples from the literature on occupational exposure and lung cancer are reported. It is demonstrated that conservative bias may decrease a numerical estimate by more than a factor 10 and that decreases by a factor 2 or 3 are not unusual. It is concluded that it is important, in future review studies, to pay attention to the magnitude of the conservative biases in the published empirical estimates and to put most emphasis on the least biased estimates in the review process.

In plain English

This single-author methodological paper used published Nordic studies of occupational lung cancer as examples. It showed how methods can shrink the 'etiological fraction': the share of disease that would not have occurred without the exposure. A Finnish study's published 2% became 56% when recalculated against a dust-free reference group. A Norwegian study's occupational estimate ranged from 22% to 47% depending on the reference group. Its asbestos estimate fell from 22% to 16% when people with uncertain, indirect or light exposure joined the comparison group. It fell to 7% when moderate exposure of over one but under ten years, or heavy exposure under a year, was also added. In a Swedish study, comparing dead cases with living controls biased estimates strongly toward null. With stated assumptions, the author reconstructed a 19% asbestos estimate. The author concluded that the asbestos fraction exceeded 5% in every Nordic study recording it separately, that most published estimates were strongly biased toward null, and that reviews should favor the best studies. It noted that counting possibly exposed people as exposed keeps estimates unbiased but raises standard error.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Methodological study: recalculation of etiological fraction estimates from published data of earlier studies
• Who was studied: Examples drawn from six longitudinal, population-based Nordic studies of occupational exposure and lung cancer (four case-control, two follow-up) from Sweden, Norway and Finland, published 1985 to 1988. Numerical recalculations used data from selected studies among these.
• Exposure: Occupational exposure in general (dusty and other occupations, respiratory carcinogens) and asbestos exposure specifically; no fiber types named
• Main results:
  • Data from Tenkanen et al. (Finland): the original estimate was 2%. Recalculated by the author against the dust-free reference group, the etiological fraction was 56%.
  • Data from Kjuus et al. (Norway): the occupational etiological fraction varied between 22% and 47% depending on the choice of reference group.
  • Asbestos data from Kjuus et al.: the etiological fraction was 22% with the original no-exposure reference group. It fell to 16% when the 'uncertain, indirect, light/sporadic exposure' group was combined into the reference group. It fell to 7% when the group with moderate exposure for more than a year but less than 10 years, or heavy exposure for less than 1 year, was also combined into it. Moderate exposure for 10 years or more, or heavy exposure for 1 year or more, stayed in the highest group.
  • Asbestos data from Kjuus et al.: the relative risk in the most exposed group changed only from 4.1 to 3.4 across the choices of reference group.
  • Asbestos data from Damber and Larsson (Sweden): reported exposure prevalence among controls was two times as high when information came from living controls rather than from next of kin of dead controls.
  • In the same Swedish data, comparing dead cases with living controls produced a strong bias toward null in both the odds ratio and the etiological fraction.
  • Using the same Swedish data and stated assumptions, the author reconstructed an asbestos-related etiological fraction of 19%.
• The authors' conclusion: The author concluded that the etiological fraction for asbestos exposure alone was above 5% in all Nordic studies that recorded asbestos separately, and that bias toward null in most published estimates was strong. The author also concluded that reviews should select the methodologically best studies rather than present the range of all estimates.
• Limitations the authors state:
  • Grouping possibly exposed persons with the exposed, so that the reference group is truly unexposed, gives an unbiased estimate but increases its standard error.
  • The 19% asbestos estimate depended on assumptions about the representativeness of living controls and the unbiasedness of the next-of-kin odds ratio.

Topics

The National Library of Medicine indexes this study under subjects our site covers: Lung cancer.

Medical terms in this study, explained in our Lexicon: Lung Cancer, Occupational Exposure.

Related studies

• Modeling Variables With a Spike at Zero: Examples and Practical Recommendations — E. Lorenz, C. Jenkner, et al. (2017)
• 'Environment' in cancer causation and etiological fraction: limitations and ambiguities — P. Boffetta, J.K. McLaughlin, et al. (2007)
• 'Environment' in cancer causation and etiological fraction: limitations and ambiguities — D. Mirabelli, E. Merler (2008)
• Accounting for outcome misclassification in estimates of the effect of occupational asbestos exposure on lung cancer death — J.K. Edwards, S.R. Cole, et al. (2014)
• Analysis of the effect of occupational exposure to asbestos based on threshold regression modeling of case-control data — A. Chambaz, D. Choudat, et al. (2014)
• The future excess fraction of occupational cancer among those exposed to carcinogens at work in Australia in 2012 — R.N. Carey, S.J. Hutchings, et al. (2017)

Browse more research in American Journal of Industrial Medicine, or return to medical research.