A Bayesian approach to strengthen inference for case-control studies with multiple error-prone exposure assessments

Our plain-English summary of A Bayesian approach to strengthen inference for case-control studies with multiple error-prone exposure assessments, by J. Zhang, S.R. Cole, D.B. Richardson, et al (Statistics in Medicine, 2013).

The study

• Authors: J. Zhang, S.R. Cole, D.B. Richardson, H. Chu
• Journal: Statistics in Medicine
• Published: 2013;32(25):4426-37
• Cited by: 8 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

In case-control studies, exposure assessments are almost always error-prone. In the absence of a gold standard, two or more assessment approaches are often used to classify people with respect to exposure. Each imperfect assessment tool may lead to misclassification of exposure assignment; the exposure misclassification may be differential with respect to case status or not; and, the errors in exposure classification under the different approaches may be independent (conditional upon the true exposure status) or not. Although methods have been proposed to study diagnostic accuracy in the absence of a gold standard, these methods are infrequently used in case-control studies to correct exposure misclassification that is simultaneously differential and dependent. In this paper, we proposed a Bayesian method to estimate the measurement-error corrected exposure-disease association, accounting for both differential and dependent misclassification. The performance of the proposed method is investigated using simulations, which show that the proposed approach works well, as well as an application to a case-control study assessing the association between asbestos exposure and mesothelioma.

Copyright © 2013 John Wiley & Sons, Ltd.

In plain English

This was a statistics methods paper. The authors proposed a Bayesian method for case-control studies that measure exposure with two or more imperfect methods. It lets measurement errors differ between cases and controls and be related across methods, and was tested in computer simulations. It was also applied to earlier case-control data on asbestos exposure and mesothelioma: 208 cases (identified 1975 to 1980) and 533 controls. Exposure, recorded as yes or no, had been assessed by next-of-kin respondents, an industrial hygienist and a job-exposure matrix. Single methods gave odds ratios of 10.74 (95% confidence limits 7.27 to 15.94) for next-of-kin, 4.65 (3.19 to 6.77) for the hygienist and 2.06 (1.48 to 2.86) for the matrix. The best-fitting of nine models gave an odds ratio of 15.46 (95% credible interval 7.88 to 52.03). Simulations showed that choosing the wrong error model could bias results substantially, in an unknown direction. The authors concluded that combining assessments gives more accurate estimates. They noted that results depend strongly on prior information and that the method handles only yes/no exposure.

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: Statistical methods paper: a new Bayesian misclassification-correction method, evaluated by computer simulation and applied as a reanalysis of data from an earlier population-based case-control study
• Who was studied: Data from an earlier population-based case-control study (Nam et al. 2005): 208 mesothelioma cases and 533 controls. Cases came from the New York State Health Department Cancer Registry, the Los Angeles County Cancer Surveillance Program and 39 Veterans Administration Hospitals, 1975 to 1980. Controls came from the same geographical area or the same hospital. The simulations used 1000 replications, each with 250 simulated cases and 500 simulated controls.
• Exposure: Asbestos exposure classified as exposed or not exposed (no fiber type given), assessed three ways: by next-of-kin respondents, by expert assessment by an industrial hygienist, and by a generic job-exposure matrix
• Main results:
  • For the asbestos–mesothelioma data, the article reports odds ratios based on single assessment methods of 10.74 (95% confidence limits 7.27, 15.94) for next-of-kin, 4.65 (3.19, 6.77) for expert assessment and 2.06 (1.48, 2.86) for the job-exposure matrix
  • Odds ratios from the nine models that used all three assessments together differed considerably from those based on any single assessment
  • Model 8, which allowed errors that were both related across methods and different between cases and controls, fit best (smallest DIC). It gave a posterior median odds ratio of 15.46 with a 95% credible interval of 7.88 to 52.03.
  • Using different prior distributions for the dependence parameters gave similar results (odds ratios from 15.32 to 16.33 in the sensitivity analysis)
  • In simulations, assuming the wrong type of misclassification could cause substantial bias in the exposure–disease association, in an unknown direction
  • In a simulation with no true association (true odds ratio 1.00), the method estimated 1.05, compared with 0.95 for next-of-kin, 0.85 for expert and 0.73 for the job-exposure matrix
• The authors' conclusion: The authors concluded that estimates combining multiple exposure assessments are more accurate than single-assessment estimates. They also concluded that wrongly specifying the misclassification model can bias results in an unpredictable direction, so competing misclassification models deserve careful attention.
• Limitations the authors state:
  • With fewer than four exposure assessments, the posterior results depend strongly on the prior information used
  • The method only handles exposure as a yes/no status, for both true and measured exposure
  • Using only 1000 simulation replications leaves some simulation error
  • For extensions to categorical exposures, informative priors for the misclassification parameters may be hard to obtain

Topics

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

Medical terms in this study, explained in our Lexicon: Case-Control Study, Asbestos Exposure, Mesothelioma.

Related studies

• Estimation and inference for case-control studies with multiple non-gold standard exposure assessments: with an occupational health application — H. Chu, S.R. Cole, et al. (2009)
• Hierarchical regression for multiple comparisons in a case-control study of occupational risks for lung cancer — M. Corbin, L. Richiardi, et al. (2012)
• A Bayesian approach to assess interaction between known risk factors: the risk of lung cancer from exposure to asbestos and smoking — D. Wraith, K. Mengersen (2008)
• Spatial variation of mortality for common and rare cancers in Piedmont, Italy, from 1980 to 2000: a Bayesian approach — M. Maule, F. Merletti, et al. (2006)
• 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)
• A quantitative weight of evidence assessment of Hill's guidelines for causal inference for cosmetic talc as a cause of mesothelioma — A.M. Ierardi, A. Urban, et al. (2021)

Browse return to medical research.