Abstract
BACKGROUND.: The detailed lung radiographic response to silica exposure has not been described. In estimating the exposure-response relationship in silicosis with statistical models, the absence of baseline (unattributable) risk can disable relative-rate estimation or produce widely varying estimates. This obstructs identification of optimum exposure metrics and invalidates comparisons and meta-analyses, which assume a common background rate. METHODS.: A cohort of 3,000 Chinese tin miners with more than 1,000 cases of silicosis was analyzed for the period 1961-1994. Regular surveillance documented three stages of silicosis. To examine the exposure-response relationship, the intercept in relative-rate models was fixed to correspond to 1% of the observed silicosis rate. Exposure metrics for contributions in different time-windows were simultaneously evaluated, as were burden and cumulative burden metrics. RESULTS.: Silica exposures that most contributed to silicosis onset occurred in the period 5-10 years prior (excess annual rate per 10 mg-year/m(3) , ER = 0.158, 95% CI = 0.125-0.192, or 16% per year). During 10-20 year prior, the excess rate contribution was much smaller (ER = 0.048, 95% CI = 0.037-0.060) but larger again during 20-30 year prior to onset (ER = 0.112, 95% CI = 0.098-0.126). For advanced silicosis, all time periods contributed about equally to the rate of onset. CONCLUSIONS.: Reliable estimates of parameters were observed, demonstrating exposure contributions over time. Burden metrics with different half-lives suggested some reversibility for silicosis onset with a half-life of 20 years. Advanced silicosis was best predicted with a cumulative burden metric which was consistent with prior observations that previously deposited silica continues to cause pulmonary damage.