Our scoring of the lung parenchymal signs was mostly based on published guidelines and reference images. There might be over- or underdetection of the abnormalities as in any such study. However, if such misclassification is not related to the exposure status (in this case the SHS exposure), it is random and leads to an underestimation of any true effects rather than to any systematic bias. The interpreters had several years of experience in assessing HRCT scans, using the window settings, of which they had had experience in clinical practice. To control for any potential bias due to the two separate image reader teams, we adjusted for this effect in the models.