Breath metabolites could reflect pathogenic events at the respiratory level and might be used to screen for the presence of occupational environmental lung diseases. The objective of this study was to estimate the accuracy of breath analysis for the diagnosis of pneumoconiosis using gas chromatography/mass spectrometry (GC/MS). A prospective study was design. We screened 200 subjects from a cohort of stone workers and their family members. Diagnosis of pneumoconiosis was made based on occupational history, the presence of abnormal pulmonary function, and parenchymal abnormalities consistent with pneumoconiosis on chest X-rays. After excluding subjects with asthma and those using steroids or taking nonsteroidal anti-inflammatory drugs on the examination day, we conducted a case-control study that enrolled 25 cases of pneumoconiosis and 154 controls. Breath and environmental air samples were collected and analysed by GC/MS. After subtraction of environmental volatile organic compounds (VOCs), the concentrations of VOCs in breath were used to discriminate between pneumoconiosis and controls. The discrimination accuracy was validated by both leave-one-out cross-validation and 5-fold cross-validation. The results showed that 94.4% of subjects were accurately classified and the cross-validated accuracy was 88.8%. The area under curve (AUC) for the diagnosis of pneumoconiosis was 0.9 (95% CI=0.8—1.0). Mean AUC of 5-fold cross-validation was 0.9. Breath test might have potential in the screening for pneumoconiosis; however, a multi-centre study is warranted to establish a reliable model and all procedures must be standardised before clinical application.