Accounting for outcome misclassification in estimates of the effect of occupational asbestos exposure on lung cancer death

Our plain-English summary of Accounting for outcome misclassification in estimates of the effect of occupational asbestos exposure on lung cancer death, by J.K. Edwards, S.R. Cole, H. Chu, et al (American Journal of Epidemiology, 2014).

The study

• Authors: J.K. Edwards, S.R. Cole, H. Chu, A.F. Olshan, D.B. Richardson
• Journal: American Journal of Epidemiology
• Published: 2014;179(5):641-7
• Cited by: 15 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

In studies of the health effects of asbestos, lung cancer death is subject to misclassification. We used modified maximum likelihood to explore the effects of outcome misclassification on the rate ratio of lung cancer death per 100 fiber-years per milliliter of cumulative asbestos exposure in a cohort study of textile workers in Charleston, South Carolina, followed from 1940 to 2001. The standard covariate-adjusted estimate of the rate ratio was 1.94 (95% confidence interval: 1.55, 2.44), and modified maximum likelihood produced similar results when we assumed that the specificity of outcome classification was 0.98. With sensitivity assumed to be 0.80 and specificity assumed to be 0.95, estimated rate ratios were further from the null and less precise (rate ratio = 2.17; 95% confidence interval: 1.59, 2.98). In the present context, standard estimates for the effect of asbestos on lung cancer death were similar to estimates accounting for the limited misclassification. However, sensitivity analysis using modified maximum likelihood was needed to verify the robustness of standard estimates, and this approach will provide unbiased estimates in settings with more misclassification.

In plain English

This statistical methods paper tested a way to correct for death certificates recording a lung cancer death as another cause, or the reverse. The authors applied it to 3,072 men and women employed at least one month between 1940 and 1965 at a South Carolina chrysotile asbestos textile plant. Follow-up to 2001 recorded 198 lung cancer deaths and 1,763 other deaths. In the standard analysis, which assumed no errors and adjusted for sex, race, age and year of study entry, the lung cancer death rate rose 1.94-fold (95% confidence interval 1.55 to 2.44) per 100 fiber-years per milliliter of cumulative exposure. Assuming imperfect certificates raised the estimates, mostly depending on specificity: how often other deaths were correctly not called lung cancer. Corrected rate ratios, per the same exposure amount, ranged from 2.00 (1.56 to 2.56) to 3.07 (1.54 to 6.10); those assuming the lowest specificity were much less precise. In simulations where error rates were assumed known, the corrected method showed little bias. The authors concluded that standard estimates were similar to estimates accounting for the limited misclassification. The study lacked smoking data and assumed dates of death were correct.

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: Methods study: a sensitivity analysis using modified maximum likelihood (Poisson regression) applied to an existing retrospective cohort, plus Monte Carlo computer simulations
• Who was studied: 3,072 men and women who had worked at least one month between 1940 and 1965 at a textile plant in Charleston, South Carolina, followed to 2001. The cohort was predominantly male (58.9%) and white (81.4%), with a median age of 23 at study entry.
• Exposure: Chrysotile asbestos at a plant making asbestos textile products. Cumulative exposure in fiber-years per milliliter was estimated from work histories with a job-exposure matrix and lagged 10 years. Median cumulative exposure at the end of follow-up was 4.99 fiber-years/mL.
• Main results:
  • The cohort had 198 lung cancer deaths and 1,763 other deaths recorded between 1940 and 2001. 265 participants (9%) were lost to follow-up.
  • Standard analysis, assuming cause of death was recorded without error: the lung cancer death rate rose by a factor of 1.94 per 100 fiber-years/mL (95% CI 1.55 to 2.44), adjusted for sex, race, age and calendar year of study entry. Table 2 gives the crude, unadjusted rate ratio as 3.52 (95% CI 2.86 to 4.33).
  • The change in the rate ratio depended mainly on the assumed specificity, not the assumed sensitivity.
  • With specificity set to 0.98, the rate ratios per 100 fiber-years/mL were 2.03 (95% CI 1.57 to 2.61), 2.02 (95% CI 1.57 to 2.60) and 2.00 (95% CI 1.56 to 2.56) for sensitivity of 0.9, 0.8 and 0.6.
  • With specificity set to 0.95, the rate ratios per 100 fiber-years/mL were 2.19 (95% CI 1.60 to 3.00), 2.17 (95% CI 1.59 to 2.98) and 2.12 (95% CI 1.56 to 2.89) for sensitivity of 0.9, 0.8 and 0.6.
  • With specificity set to 0.90, the estimates were further from the null and much less precise. Table 2 gives 2.97 (95% CI 1.34 to 6.56), 3.03 (95% CI 1.32 to 6.94) and 3.07 (95% CI 1.54 to 6.10) for sensitivity of 0.9, 0.8 and 0.6.
  • The authors ran simulations on 10,000 simulated cohorts. Standard estimates were biased toward the null when certificates were imperfect. Estimates from the modified method, using sensitivity and specificity values assumed to be known, showed little bias.
• The authors' conclusion: The authors concluded that, in this cohort, standard estimates of the effect of asbestos on lung cancer death were similar to estimates accounting for the limited misclassification. They added that standard methods are likely to give notably biased estimates for outcomes with more misclassification, particularly poor specificity.
• Limitations the authors state:
  • The study did not collect information on smoking history.
  • Few validation studies reported the specificity of death certificates for lung cancer, so the authors tested a range of assumed values (0.98, 0.95, 0.90).
  • The analysis assumed misclassification did not differ by exposure level. The authors note that lung cancer might be suspected earlier in heavily exposed people.
  • The analysis assumed dates of death were correct and only the cause of death could be wrong. The method would not be enough if death dates or deaths themselves were misrecorded.

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: Asbestos Exposure, Lung Cancer, rate ratio, Fiber Year, Cohort Study, confidence interval.

Related studies

• Analysis of occupational asbestos exposure and lung cancer mortality using the g formula — S.R. Cole, D.B. Richardson, et al. (2013)
• Estimating the effect of cumulative occupational asbestos exposure on time to lung cancer mortality: using structural nested failure-time models to account for healthy-worker survivor bias — A.I. Naimi, S.R. Cole, et al. (2014)
• Unbiased vs. conservative estimators of etiological fractions: examples of misclassification from studies of occupational lung cancer — O. Olsen (1995)
• Assessing the component associations of the healthy worker survivor bias: occupational asbestos exposure and lung cancer mortality — A.I. Naimi, S.R. Cole, et al. (2013)
• An estimate of cancers attributable to occupational exposures in France — P. Boffetta, P. Autier, et al. (2010)
• Estimates of historical exposures by phase contrast and transmission electron microscopy for pooled exposure--response analyses of North Carolina and South Carolina, USA asbestos textile cohorts — J.M. Dement, D. Loomis, et al. (2011)

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