Analysis of the effect of occupational exposure to asbestos based on threshold regression modeling of case-control data

Our plain-English summary of Analysis of the effect of occupational exposure to asbestos based on threshold regression modeling of case-control data, by A. Chambaz, D. Choudat, C. Huber, et al (Biostatistics, 2014).

The study

• Authors: A. Chambaz, D. Choudat, C. Huber, J.C. Pairon, M.J. van der Laan
• Journal: Biostatistics
• Published: 2014;15(2):327-40
• Read it: PubMed · Publisher

Abstract

We analyze the effect of occupational exposure to asbestos on the occurrence of lung cancer based on a recent French case-control (CC) study. We build a large collection of threshold regression models, data-adaptively select a better model by CC-weighted likelihood-based cross-validation and then fit this better model by CC-weighted maximum likelihood. The CC-weighting allows us to draw valid inferences from CC data without relying on a logistic regression. This is possible because the joint distribution of the indicator of being a case and matching variable is available beforehand owing to two studies independent from our data set. The implications of the fitted model in terms of years of life free of lung cancer lost due to the exposure to asbestos are discussed.

In plain English

This was a statistical methods paper that re-analyzed data from a French matched case-control study of lung cancer (Pairon and others, 2009), carried out in four Paris hospitals between 1999 and 2002. After people with missing information were excluded, the data covered 860 lung cancer cases and 901 controls. Each job held for at least six months was rated for the probability, frequency and intensity of occupational asbestos exposure. The article does not name fiber types. The authors used a 'threshold regression' model in which asbestos exposure speeds up a person's biological clock toward lung cancer. They chose among many versions of the model by cross-validation and estimated years of life free of lung cancer lost. In the chosen model, passive exposure had the same effect as no exposure. For 601 cases the estimated years lost were 0. Among the other 259 cases (30% of cases), the median was 0.769 years and the mean 2.467 years, with 95% lower bounds of 0.037 and 0.555. The authors noted limitations, including the lack of a detailed smoking history.

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 modeling study (threshold regression with case-control weighting and cross-validated model selection) applied to existing matched case-control data
• Who was studied: Patients of four Parisian hospitals, 1999–2002: 860 lung cancer cases and 901 controls (1761 observations) after excluding subjects with missing information; controls came from ophthalmology, general surgery and orthopedic surgery departments
• Exposure: Occupational asbestos exposure for each job lasting at least 6 months, rated by experts for probability, frequency and intensity (each low, mild or high), plus a no-exposure category, giving 28 categories. Intensity levels corresponded to less than 0.1, 0.1 to 1, and more than 1 asbestos fibers per mL. Fiber types were not specified.
• Main results:
  • In the selected model, passive exposure (probability index 1) had the same effect as no exposure, and frequency levels 1 and 2 had the same effect
  • For 601 of the 860 cases, estimated years of life free of lung cancer lost were exactly 0, because they were either not exposed or only passively exposed
  • Among the 259 cases (30% of all cases) with positive estimated years lost: minimum 0.026, 25th percentile 0.289, median 0.769, mean 2.467, 75th percentile 2.408, maximum 36.577 years
  • Corresponding 95% confidence lower bounds for those 259 cases: minimum 0.001, 25th percentile 0.014, median 0.037, mean 0.555, 75th percentile 0.102, maximum 12.832 years
  • For a man and a woman with the same asbestos history and smoking level, the model indicated the man was more likely to have developed lung cancer by a given age
  • For two people of the same gender and asbestos history, the one with the larger lifetime tobacco use was more likely to have developed lung cancer by a given age
  • Occurrence of lung cancer in close family did not enter the selected model
  • The estimated exposure weights differed greatly from the values chosen arbitrarily in the original Pairon and others (2009) analysis
• The authors' conclusion: The authors stated that their weighting approach allowed valid inferences from case-control data without relying on logistic regression. They said they intend to extend the approach in future work.
• Limitations the authors state:
  • The model ignored how long people smoked and years since quitting, although these are known to affect lung cancer
  • No year-by-year smoking history was available, which the authors believed could cause residual confounding
  • The original matching pattern and race were not available, so an artificial matching pattern was used
  • The 28-level exposure description was too detailed, difficult to establish and prone to errors
  • The analysis assumed that national French cancer data are representative of the Parisian study population
  • The authors assumed that truncation could be neglected, even though cases were not interviewed immediately on diagnosis
  • The theoretical guarantee motivating the model-selection procedure could not be rigorously applied here
  • Years of life lost is not necessarily a remaining number of years, since death may occur at any time after lung cancer even without exposure

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: Occupational Exposure, Lung Cancer.

Related studies

• Hierarchical regression for multiple comparisons in a case-control study of occupational risks for lung cancer — M. Corbin, L. Richiardi, et al. (2012)
• Modeling Variables With a Spike at Zero: Examples and Practical Recommendations — E. Lorenz, C. Jenkner, et al. (2017)
• Estimating the potential disease rate reduction under varying conditions of combined effects — H. Checkoway, J.L. Hickey (1988)
• Mapping cancer risk in southwestern Ontario with changing census boundaries — P. Nguyen, P.E. Brown, et al. (2012)
• A case-control study relating railroad worker mortality to diesel exhaust exposure using a threshold regression model — M.L. Lee, G.A. Whitmore, et al. (2009)
• 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)

Browse return to medical research.