Hierarchical latency models for dose-time-response associations

Our plain-English summary of Hierarchical latency models for dose-time-response associations, by D.B. Richardson, R.F. MacLehose, B. Langholz, et al (American Journal of Epidemiology, 2011).

The study

• Authors: D.B. Richardson, R.F. MacLehose, B. Langholz, S.R. Cole
• Journal: American Journal of Epidemiology
• Published: 2011;173(6):695-702
• Cited by: 17 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Exposure lagging and exposure-time window analysis are 2 widely used approaches to allow for induction and latency periods in analyses of exposure-disease associations. Exposure lagging implies a strong parametric assumption about the temporal evolution of the exposure-disease association. An exposure-time window analysis allows for a more flexible description of temporal variation in exposure effects but may result in unstable risk estimates that are sensitive to how windows are defined. The authors describe a hierarchical regression approach that combines time window analysis with a parametric latency model. They illustrate this approach using data from 2 occupational cohort studies: studies of lung cancer mortality among 1) asbestos textile workers and 2) uranium miners. For each cohort, an exposure-time window analysis was compared with a hierarchical regression analysis with shrinkage toward a simpler, second-stage parametric latency model. In each cohort analysis, there is substantial stability gained in time window-specific estimates of association by using a hierarchical regression approach. The proposed hierarchical regression model couples a time window analysis with a parametric latency model; this approach provides a way to stabilize risk estimates derived from a time window analysis and a way to reduce bias arising from misspecification of a parametric latency model.

© The Author 2011. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved.

In plain English

This methods paper proposed a statistical approach, hierarchical regression, for studying how an exposure-disease link changes with time since exposure. It coupled a time-window analysis with a simpler latency model, such as a lag or a rise-then-fall pattern, letting window estimates depart from that model when data warranted. The authors illustrated it with two earlier occupational studies. The asbestos example used South Carolina asbestos textile workers employed between 1940 and 1965. Chrysotile exposure was measured in fiber-years per mL, and each of 198 lung cancer deaths was compared with up to 40 controls. In a single-stage model with all eight time windows, the estimates were highly imprecise. Under a 10-year lag, the excess relative risk of lung cancer death per 100 fiber-years per mL was 2.31 (95% credible interval 1.18 to 3.53). The hierarchical model gave 2.43 (0.96 to 4.09), with substantially less variable window estimates. In the second example, radon-exposed uranium miners, a rise-then-fall pattern fit better than a simple lag. The authors concluded the approach can stabilize time-window estimates and reduce bias from a misspecified latency model.

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 proposing a hierarchical (two-stage) regression model for latency analysis, illustrated with nested case-control analyses of two existing occupational cohorts
• Who was studied: Asbestos example: workers employed in asbestos textile production at a plant in South Carolina (Charleston) between January 1, 1940 and December 31, 1965, followed for vital status through December 31, 2001. There were 198 lung cancer deaths, each with up to 40 controls. Second example: 263 lung cancer deaths among men first employed in underground uranium mining in the Colorado Plateau, 1950 to 1960.
• Exposure: Cumulative chrysotile asbestos exposure (fibers longer than 5 micrometers per mL of air, summed over days worked, in fiber-years per mL), examined in 8 time windows since exposure (0 to under 5 years up to 35 or more years). Radon exposure in working-level months for the uranium miners.
• Main results:
  • Asbestos textile workers, single-stage model with 8 parameters for the 8 time windows: the estimates were highly imprecise for every time window. The coefficient was negative for the 5 to under 10 years window and positive for the others.
  • Asbestos textile workers, cumulative exposure under a 10-year lag: excess relative risk of lung cancer mortality per 100 fiber-years per mL was 2.31 (95% credible interval 1.18, 3.53).
  • Asbestos textile workers, hierarchical model: the time window-specific estimates varied substantially less. The second-stage estimate was an excess relative risk per 100 fiber-years per mL of 2.43 (95% credible interval 0.96, 4.09).
  • Asbestos textile workers: the model-fit criterion (DIC) was 1,396.6 for the hierarchical model and 1,402.8 for the single-stage time-window model. The authors said this supports preferring the hierarchical model.
  • Uranium miners: a 10-year lag did not describe the radon data well. With a bilinear (rise-then-fall) second-stage model, the association peaked at 10.2 years after exposure (95% credible interval 5.0, 16.1), with no excess effect more than 36.2 years after exposure. The maximal excess relative risk per 1,000 working-level months was 7.06 (95% credible interval 2.85, 11.94).
• The authors' conclusion: The authors concluded that the hierarchical model, which couples a time-window analysis with a parametric latency model, provides a way to stabilize risk estimates from time-window analyses and to reduce bias from misspecifying a parametric latency model.

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: Latency, Lung Cancer.

Related studies

• 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)
• Dose-response in case-control studies — G. Berry (1980)
• Dose-response associations of silica with nonmalignant respiratory disease and lung cancer mortality in the diatomaceous earth industry — H. Checkoway, N.J. Heyer, et al. (1997)
• Re: "Dose-response associations of silica with nonmalignant respiratory disease and lung cancer mortality in the diatomaceous earth industry" — G.W. Gibbs (1998)
• Re: "Dose-response associations of silica with nonmalignant respiratory disease and lung cancer mortality in the diatomaceous earth industry" — M.J. Mirliss (1998)
• Pleural plaques and asbestosis: dose- and time-response relationships based on HRCT data — C. Paris, S. Thierry, et al. (2009)

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