Objectives Despite the scale of the silicosis problem in the South African goldmining industry, studies in active miners are scarce. A large randomised cluster trial of isoniazid tuberculosis prophylaxis provided an opportunity to investigate the predictors of silicosis and variation in silicosis prevalence across goldmining shafts.

Methods This analysis was based on a random sample of routine annual chest radiographs of active miners entering the trial, taken between 2004 and 2009. All were read for silicosis using the ILO classification by an experienced lay reader. All films classified as abnormal and a random sample of normals were re-read by a ‘B’-reader and these readings substituted for those of the lay reader in case of disagreement. The association between silicosis defined as ILO > 1/0 and age at radiograph, years since first employment, shaft and other occupational markers, was analysed.

Results A total of 14,434 radiographs from 15 goldmining shafts were read. Silicosis prevalence varied from 0.68% to 6.28% across shafts; 7.8 fold across the range in adjusted analysis. Silicosis showed a strong dose response relationship with years since first employment [OR 8.1 (95% CI 3.2, 21) for > 30 vs < 15 years]. Surprisingly, after adjusting for years since first employment, age was still predictive of silicosis risk (p < 0.001). Miners from Lesotho, who make up a large minority of the workforce, had an increased odds (OR 1.55; 95% CI 1.2, 1.9) of silicosis relative to South African miners.

Conclusions Goldmining shafts differ in silicosis risk independently of workforce age and length of mining service. Speculative explanations include higher silica fraction of the rock and greater intensity of exposure to silica. Basotho miners are at increased relative risk of silicosis, suggesting greater involvement in dustier jobs. The strong independent effect of age on silicosis risk, reported previously, remains to be explained.