Disease mapping of zero-excessive mesothelioma data in Flanders

Our plain-English summary of Disease mapping of zero-excessive mesothelioma data in Flanders, by T. Neyens, A.B. Lawson, R.S. Kirby, et al (Annals of Epidemiology, 2017).

The study

• Authors: T. Neyens, A.B. Lawson, R.S. Kirby, V. Nuyts, K. Watjou, M. Aregay, R. Carroll, T.S. Nawrot, C. Faes
• Journal: Annals of Epidemiology
• Published: 2017;27(1):59-66.e3
• Cited by: 12 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Purpose: To investigate the distribution of mesothelioma in Flanders using Bayesian disease mapping models that account for both an excess of zeros and overdispersion.

Methods: The numbers of newly diagnosed mesothelioma cases within all Flemish municipalities between 1999 and 2008 were obtained from the Belgian Cancer Registry. To deal with overdispersion, zero inflation, and geographical association, the hurdle combined model was proposed, which has three components: a Bernoulli zero-inflation mixture component to account for excess zeros, a gamma random effect to adjust for overdispersion, and a normal conditional autoregressive random effect to attribute spatial association. This model was compared with other existing methods in literature.

Results: The results indicate that hurdle models with a random effects term accounting for extra variance in the Bernoulli zero-inflation component fit the data better than hurdle models that do not take overdispersion in the occurrence of zeros into account. Furthermore, traditional models that do not take into account excessive zeros but contain at least one random effects term that models extra variance in the counts have better fits compared to their hurdle counterparts. In other words, the extra variability, due to an excess of zeros, can be accommodated by spatially structured and/or unstructured random effects in a Poisson model such that the hurdle mixture model is not necessary.

Conclusions: Models taking into account zero inflation do not always provide better fits to data with excessive zeros than less complex models. In this study, a simple conditional autoregressive model identified a cluster in mesothelioma cases near a former asbestos processing plant (Kapelle-op-den-Bos). This observation is likely linked with historical local asbestos exposures. Future research will clarify this.

Copyright © 2016 Elsevier Inc. All rights reserved.

In plain English

In this statistical methods study, the authors analysed Belgian Cancer Registry data on newly diagnosed mesothelioma (pleura, peritoneum and pericardium) in the 308 municipalities of Flanders, excluding Brussels, from 1999 to 2008. The data had one count per municipality, year and sex (6,160 counts), and because the disease is rare, 84% of the counts were zero. The authors compared Bayesian disease-mapping models, including "hurdle" models with a separate part for excess zeros. Among hurdle models, those allowing extra variation in the zero part fitted better. Traditional models without that separate part fitted better still, and a simple spatial (CAR) model fitted best. The authors explained that the random effects in these models could absorb the extra variation from the many zeros. On the CAR model's map, relative risks were strongly elevated in and around Kapelle-op-den-Bos, where the asbestos cement maker Eternit NV used asbestos until 1994. Central Flanders generally had higher risks than the east and west. The models included no exposure information. The authors called the Kapelle-op-den-Bos result convincing given Eternit's asbestos use. They cautioned that cases were counted where people were diagnosed, not necessarily where they were exposed, and that risks were assumed constant over time and sex.

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: Geographical (disease mapping) analysis of cancer registry counts, comparing Bayesian statistical models, including hurdle models for excess zeros
• Who was studied: Newly diagnosed mesothelioma cases (pleura, peritoneum and pericardium) in males and females in the 308 municipalities of Flanders (Belgium, without Brussels), 1999 to 2008, from the Belgian Cancer Registry. There was one count per municipality, year and sex, 6160 observations in all.
• Exposure: No individual exposure was measured. Location was analysed, with attention to Kapelle-op-den-Bos, home of Eternit NV, an asbestos cement manufacturing company that used asbestos until 1994. Fiber types are not named.
• Main results:
  • 84% of the counts were zero. The data were counts per municipality, year and sex, 6160 observations in all, and the models predicted a probability of zeros of about 0.84.
  • Among hurdle models, those with a random effect allowing extra variation in the zero (Bernoulli) component fitted better than those assuming constant zero-inflation
  • Hurdle models fitted worse than the traditional models, which have no separate zero component. For example, the hurdle combined model (M = -3133.14) fitted worse than the traditional combined model (M = -2796.93), where a larger M is better.
  • The authors explained that random effects in the traditional models could absorb the extra variation caused by the many zeros
  • The best-fitting model was the CAR model without an unstructured term (M = -2789.75), almost equal to the CAR convolution model (M = -2789.84)
  • On the relative risk map from the CAR model, risks were strongly elevated in and around Kapelle-op-den-Bos. The central part of Flanders generally had increased risks compared with the eastern and western parts.
  • When males and females were modeled separately, relative risk estimates were not significantly different between the sexes in more than 95% of the municipalities
• The authors' conclusion: The authors concluded that the data were modeled best by a traditional model rather than a hurdle model for excess zeros. They found elevated relative risks in central Flanders, especially in Kapelle-op-den-Bos and its north-western neighbouring municipalities. Because Kapelle-op-den-Bos is known for Eternit NV and its use of asbestos, they called this a convincing result.
• Limitations the authors state:
  • Counts reflect where patients were diagnosed, not necessarily where they were exposed to asbestos, especially given the long incubation time
  • Relative risks were assumed constant over time and sex, which the authors call a strong assumption made to simplify the model
  • Indirect gender-age standardization may be a source of bias, and the authors warn against over-interpreting the results
  • No covariates such as year and gender were included in the models
  • The 10-year time frame may not be enough to capture longitudinal patterns
  • Other areas in Flanders may also have had historical asbestos exposure that is not clear from the map. Information on all exposure sources was not collected.

Topics

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

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

Related studies

• Impact of malignant mesothelioma in Taiwan: a 27-year review of population-based cancer registry data — L.J. Lee, Y.Y. Chang, et al. (2010)
• Management of malignant pleural mesothelioma - part 3 : Data from the Austrian Mesothelioma Interest Group (AMIG) database — T. Klikovits, M.A. Hoda, et al. (2016)
• Analysis of survival of mesothelioma cases in the Italian register (ReNaM) — A. Marinaccio, M. Nesti, et al. (2003)
• Patterns in the incidence, mortality and survival of malignant pleural and peritoneal mesothelioma, New South Wales, 1972-2009 — M.J. Soeberg, N. Creighton, et al. (2016)
• Malignant mesothelioma in Italy, 1997 — M. Nesti, A. Marinaccio, et al. (2004)
• Filing for workers' compensation among Ontario cases of mesothelioma — J.I. Payne, E. Pichora (2009)

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