Is exposure to silica associated with lung cancer in the absence of silicosis? A meta-analytical approach to an important public health question

Abstract

This report investigates epidemiologically whether exposure to silica is associated with lung cancer risks in individuals without silicosis. We searched the PubMed reference data base from 1966 through 1/2007 for reports of lung cancer in silica-exposed persons without and with silicosis. To explore heterogeneity between studies, a multi-stage strategy was employed. First, fixed-effect summaries (FES) and corresponding 95% confidence intervals (CI) for various combinations of studies were calculated, weighting individual results by their precision. The homogeneity of the contributing results was examined using chi(2) statistics. Where there was evidence of substantial heterogeneity, the CI around the FES was increased to take account of the between-study variability. Random-effect summaries and their CI for identical combinations of studies were also computed. Meta regression was used to explore interactions with covariates. To draw comparisons, parallel analyses were performed for non-silicotics and for silicotics. The persistence of a significant link between silicosis and lung cancer since the characterisation in 1997 of silica as a human carcinogen [our estimates of lung cancer relative risks (RR) exceeded unity in each of 38 eligible studies of silicotics published until 1/2007, averaging 2.1 in analyses based on both fixed and random effect models (95% CI = (2.0-2.3) and (1.9-2.3), respectively)] does not resolve our study question, namely whether exposure to silica levels below those required to induce silicosis are carcinogenic. Importantly, our detailed examination of 11 studies of lung cancer in silica-exposed individuals without silicosis included only three with data allowing adjustment for smoking habits. They yielded a pooled RR estimate of 1.0 [95% CI = (0.8-1.3)]. The other eight studies, with no adjustment for smoking habits, suggested a marginally elevated risk of lung cancer [RR = 1.2; 95% CI (1.1-1.4)], but with significant heterogeneity between studies (P approximately 0.05). Necessary further research should concentrate on silica exposures both above and below those that induce silicosis, so that the shape of the exposure-response relationship may be identified, with adjustments for likely confounding factors including silicosis. Time-dependent information on silicosis and on silica dust is required as well as the application of methods like G-estimation to answer the important public health question: Is silicosis a necessary condition for the elevation of silica-associated lung cancer risks?