Potential pitfall in using cumulative exposure in exposure-response relationships: demonstration and discussion

Our plain-English summary of Potential pitfall in using cumulative exposure in exposure-response relationships: demonstration and discussion, by M.M. Finkelstein (American Journal of Industrial Medicine, 1995).

The study

• Author: M.M. Finkelstein
• Journal: American Journal of Industrial Medicine
• Published: 1995;28(1):41-7
• Type of article: Randomized trial
• Cited by: 8 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

Cumulative exposure is frequently used as a measure of exposure in the quantitative analysis of epidemiologic studies. It is recognized that the imposed symmetry between duration and intensity of exposure is a potential problem with this measure, but it is less widely recognized that the finding of an exposure-response relationship, using cumulative exposure as the exposure metric, does not necessarily imply that exposures were accurately or even consistently estimated. This report describes a simulation study drawn from a nested case-control analysis of mesothelioma in a cohort of asbestos cement workers. Intensity of exposure in the range of 0.1-40 fibers/ml was randomly assigned to subjects. Logistic regression analysis demonstrated that there was no association between mesothelioma risk and the randomly assigned intensity of exposure. However, in 171 (86%) of 200 trials, mesothelioma risk was significantly associated with cumulative exposure, even though intensity of exposure remained randomly assigned. A strong exposure-response relationship might thus be misleading. One would be more confident about quantitative risk assessment when there are a large number of independent studies available for analysis.

In plain English

This article reported a simulation study by the author. It was built on data from an earlier nested case-control analysis of mesothelioma in a cohort of asbestos-cement factory workers. The cases were 45 men who developed mesothelioma (31 pleural and 14 peritoneal). Each was matched with five controls who were born in the same year and who survived him. The author kept each man's real duration of exposure and time since first exposure but assigned exposure intensity at random, between 0.1 and 40 fibers per ml. The analysis was repeated in 200 trials using conditional logistic regression. The randomly assigned intensity was statistically significant in 11 of 200 trials, consistent with the 5% significance level, and showed no association with mesothelioma risk. When that random intensity was combined with real duration to form cumulative exposure, it was significantly associated with mesothelioma risk in 171 (86%) of 200 trials, always pointing toward rising risk. The author concluded that a strong exposure-response relationship might be misleading, and that one would be more confident about quantitative risk assessment when many independent studies are available.

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: Simulation study (200 repeated trials) using data from a nested case-control analysis, analysed by conditional logistic regression
• Who was studied: 45 men from a cohort of asbestos-cement factory workers who developed mesothelioma (31 pleural, 14 peritoneal), each matched with five controls born in the same year who survived him. Subjects were taken from the author's earlier study (Finkelstein, 1991).
• Exposure: Asbestos exposure in asbestos-cement factory work; no fiber type is named in this article. Exposure intensity was randomly assigned in the range 0.1 to 40 fibers per ml, while the real duration of exposure and latency were taken from the actual data set.
• Main results:
  • Randomly assigned exposure intensity: the regression coefficient was statistically significant in 11 of 200 trials, consistent with the 5% significance level. The results were spread evenly around 0, meaning no association between mesothelioma risk and the randomly assigned intensity.
  • Cumulative exposure (random intensity multiplied by real duration): the coefficient was statistically significant in 171 (86%) of 200 trials and was positive in all trials. The mean coefficient over 200 trials was 1.16.
  • In most trials the relationship between cumulative exposure and risk was compatible with a linear exposure-response relationship, because the 95% confidence interval for the coefficient included 1.0.
  • In the model with intensity as a separate term, the length of exposure term was highly significant. The author noted that length of exposure is a surrogate for asbestos dose in this workplace.
• The authors' conclusion: The author concluded that a strong exposure-response relationship based on cumulative exposure might be misleading, because duration of exposure may dominate the exposure term even when intensity is randomized and largely misclassified. More confidence in quantitative risk assessment would come from a large number of independent studies.

Topics

Medical terms in this study, explained in our Lexicon: Cumulative Exposure, Work Practice Simulation, Mesothelioma, Asbestos-Cement.

Related studies

• Lagging exposure information in cumulative exposure-response analyses — D.B. Richardson, S.R. Cole, et al. (2011)
• Lung cancer risk at low cumulative asbestos exposure: meta-regression of the exposure-response relationship — S. van der Bij, H. Koffijberg, et al. (2013)
• A meta-analysis of asbestos and lung cancer: is better quality exposure assessment associated with steeper slopes of the exposure-response relationships? — V. Lenters, R. Vermeulen, et al. (2011)
• Hierarchical latency models for dose-time-response associations — D.B. Richardson, R.F. MacLehose, et al. (2011)
• The potential impact of epidemiology on the prevention of occupational disease — D.H. Wegman (1992)
• Sources of uncertainty in dose-response modeling of epidemiological data for cancer risk assessment — L. Stayner, A.J. Bailer, et al. (1999)

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