Projection of mesothelioma mortality in Britain using Bayesian methods

Our plain-English summary of Projection of mesothelioma mortality in Britain using Bayesian methods, by E. Tan, N. Warren, A.J. Darnton, et al (British Journal of Cancer, 2010).

The study

• Authors: E. Tan, N. Warren, A.J. Darnton, J.T. Hodgson
• Journal: British Journal of Cancer
• Published: 2010;103(3):430-6
• Cited by: 43 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Background: Mesothelioma mortality has increased more than ten-fold over the past 40 years in Great Britain, with >1700 male deaths recorded in the British mesothelioma register in 2006. Annual mesothelioma deaths now account for >1% of all cancer deaths. A Poisson regression model based on a previous work by Hodgson et al has been fitted, which has allowed informed statistical inferences about model parameters and predictions of future mesothelioma mortality to be made.

Methods: In the Poisson regression model, the mesothelioma risk of an individual depends on the average collective asbestos dose for the individual in a given year and an age-specific exposure potential. The model has been fitted to the data within a Bayesian framework using the Metropolis-Hastings algorithm, a Markov Chain Monte Carlo technique, providing credible intervals for model parameters as well as prediction intervals for the number of future cases of mortality.

Results: Males were most likely to have been exposed to asbestos between the ages of 30 and 49 years, with the peak year of asbestos exposure estimated to be 1963. The estimated number of background cases was 1.08 cases per million population.

Conclusion: Mortality among males is predicted to peak at approximately 2040 deaths in the year 2016, with a rapid decline thereafter. Approximately 91,000 deaths are predicted to occur from 1968 to 2050 with around 61,000 of these occurring from 2007 onwards.

In plain English

This was a statistical modelling study that projected future mesothelioma deaths among men in Great Britain. The authors refitted a modified version of an earlier model (Hodgson et al, 2005) to deaths of men aged 20–89 recorded in the British Mesothelioma Register from 1968 to 2006. They used Bayesian methods, which allowed uncertainty ranges to be calculated. In the model, risk depended on estimated yearly population asbestos exposure and on how likely men of each age were to be exposed. The model estimated that exposure peaked in 1963 and was highest among men aged 30 to 49. The background rate of mesothelioma not caused by asbestos was estimated at 1.08 cases per million, about 23 cases in 2006. Deaths among all men were projected to peak at 2038 (90% prediction interval 1929–2156) in 2016, then decline rapidly. About 91,000 deaths were projected from 1968 to 2050, around 61,000 of them from 2007 onwards. The authors concluded that the model fitted the data well and suited shorter-term projections. They noted that its ranges left out uncertainty about exposure after 1978, and that it was much less clear whether it suited longer-term projections.

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 modelling study: a Poisson regression model of mesothelioma mortality refitted with Bayesian methods (Metropolis–Hastings Markov Chain Monte Carlo), then used to project future deaths
• Who was studied: Mesothelioma deaths of males aged 20–89 years in Great Britain from 1968 to 2006, taken from the British Mesothelioma Register. Projections to 2050 used Office for National Statistics population projections.
• Exposure: Estimated average collective (population-level) asbestos exposure for British males in each year, weighted by an age-specific exposure potential. No fiber types were specified.
• Main results:
  • Estimated population asbestos exposure rose rapidly from the 1940s to the mid-1960s, reached a maximum in 1963, and decreased thereafter.
  • Estimated exposure potential was highest among males aged 30 to 49, and lowest among males under 15 and over 50.
  • Compared with males aged 20–29 (baseline, 1.00), the estimated relative exposure potential for males aged 30–39 was 1.79 (90% CI 1.51 to 2.03).
  • Compared with males aged 20–29 (baseline, 1.00), the estimated relative exposure potential for males aged 40–49 was 1.59 (90% CI 1.25 to 1.94).
  • The background rate of mesothelioma not caused by asbestos was estimated at 1.08 cases per million, about 23 cases in 2006 among males aged 20 to 89.
  • Projected peak in deaths among all males: 2038 deaths (90% prediction interval 1929–2156) in 2016 (90% CI 2015–2016), with a rapid decline after the peak year.
  • About 91,000 male deaths were projected from 1968 to 2050, around 61,000 of them from 2007 onwards.
  • Of 126 deviance residuals, 121 (96%) lay in the range minus 2 to 2, which the authors took to suggest a satisfactory fit.
  • In a sensitivity analysis, lag times of 0 and 15 years shifted the projected peak, but both fitted the data much worse than the assumed 10-year lag.
  • Including a term for changing completeness of diagnosis over time did not improve the model's fit.
• The authors' conclusion: The authors concluded that the modified model fitted mesothelioma data from 1968 to 2006 well, providing a sound basis for shorter-term projections of male mesothelioma mortality.
• Limitations the authors state:
  • The prediction intervals reflect only uncertainty in the fitted model parameters, not the potentially considerable uncertainty from the assumed shape of the exposure curve after 1978. If exposure had levelled off instead, total deaths to 2050 would be much larger.
  • It is much less clear whether the model provides a good basis for longer-term projections.
  • The lung clearance half-life could not be estimated. It was fixed at 1,000,000 years, a value the authors say should be interpreted with caution.
  • The fixed 10-year lag between exposure and disease is a strong assumption.
  • The estimated exposure profile from the mid-1960s onwards is more uncertain, because of mesothelioma's long latency and because the mortality peak had not yet been reached.

Topics

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

Medical terms in this study, explained in our Lexicon: Mesothelioma, Asbestos Exposure.

Related studies

• Trend and projection of larynx cancer incidence and mortality in China from 1990 to 2044: A Bayesian age-period-cohort modeling study — E. Ye, J. Huang, et al. (2023)
• Simple benchmark for mesothelioma projection for Great Britain — M.D. Martínez-Miranda, B. Nielsen, et al. (2016)
• The expected burden of mesothelioma mortality in Great Britain from 2002 to 2050 — J.T. Hodgson, D.M. McElvenny, et al. (2005)
• The effect of smoking on the risk of lung cancer mortality for asbestos workers in Great Britain (1971-2005) — G. Frost, A. Darnton, et al. (2011)
• Incidence of mesothelioma in Lombardy, Italy: exposure to asbestos, time patterns and future projections — C. Mensi, S. De Matteis, et al. (2016)
• Mesothelioma mortality in Great Britain from 1968 to 2001 — D.M. McElvenny, A.J. Darnton, et al. (2005)

Browse more research in British Journal of Cancer, or return to medical research.