We also compared blue-collar construction workers
to blue-collar workers in other industries. Whereas the
analysis in the entire population simulates the paradigm
of a cohort study in which the disease occurrence of
the cohort of construction workers is compared with
that of the general population, the analysis restricted to
blue-collar workers represents the situation, difficult to
achieve in a cohort study, where we lessen the opportunity
for confounding by blue-collar/white collar status and its
correlates. The resulting ORs were very similar to those
using the entire study population as reference group.
Nor was this finding sensitive to the particular cutpoint
we used to define a worker as blue-collar or not.
Results based on self-respondents were similar to the
main findings and adjustment for different combinations
of covariates had little impact.
Some previous studies indicated stronger associations
between work in construction industry and lung cancer
than what we found. There may be more mobility
between work in this industry and others in our
study area than elsewhere, as evidenced by the fact
that only 11 % of construction industry workers had
never worked in any other industry, and on average,
workers who had ever worked in this industry, spent
less than half of their careers in this industry. Still, it
is not implausible that workers in this industry as a
whole experience only a slight (11 %) excess risk of
lung cancer.
Our analyses of exposure agents (Table 3) were restricted
to construction workers and therefore entailed rather small
numbers and imprecise risk estimates. Whether to conduct
such an analysis in the entire population or to restrict the
study base to construction workers is not self-evident;
each strategy presents advantages and disadvantages.
Conducting the analysis in the entire population is the
most common approach; it maximizes numbers of subjects