Activity Number:
|
428
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Epidemiology
|
Abstract - #307203 |
Title:
|
Determining an Optimal Exposure Metric from a Bivariate Distribution of Asbestos Exposures (Length, Diameter) in a Cohort of South Carolina Textile Workers
|
Author(s):
|
Stephen Gilbert*+ and Leslie T. Stayner and Eileen D. Kuempel and John D. Dement
|
Companies:
|
National Institute for Occupational Safety and Health and University of Illinois at Chicago and National Institute for Occupational Safety and Health and Duke University Medical Center
|
Address:
|
4676 Columbia Parkway, Cincinnati, OH, 45226,
|
Keywords:
|
simulation ; epidemiology ; public health ; risk assessment ; modeling
|
Abstract:
|
What combination of asbestos fiber length/diameter cells best predicts human lung cancer? Original samples from a SC textile plant were re-analyzed using TEM to obtain more detailed bivariate (length, diameter) asbestos fiber size distributions by department and job. Separate Cox models were fit to cumulative exposure for all 36 length/diameter cells and other combinations of those 36 cells. While nearly all exposure cells and chosen combinations showed significant dose-response relationships, the fits to metrics with fibers >40µm long provide the best fits. Since exposure occurs in nearly all 36 cells we wanted to find the optimal combination of those 36 cells. So, single exposure metrics were computed as weighted sums of the 36 cells using random weights and simulated annealing. The simulation results support the notion that metrics that include long fibers provide the best fits.
|