Radiologic Detection of Pleural Thickening

The purpose of this study was to investigate four aspects of the radiologic detection of pleural thickening: the specificity of chest X-rays read by B readers, the effects of threshold definitions of positivity, the extent of variability among different X-ray readers, and the effect of film quality. A series of 421 chest X-rays corresponding to consecutive autopsies was reviewed by six B readers working independently and using modified International Labour Organization forms. The prevalence of true pleural thickening was approximately 6%. Using a strict definition of positivity, the average specificity was 0.85. Specificity decreased when a more liberal definition was used. (The average sensitivity was 0.32, which appeared to increase with a more liberal definition of positivity; because of small numbers of true positives, the sensitivity estimates were unstable.) Kappa statistics for interreader agreement ranged from 0.06 to 0.40, suggesting considerable interreader variability. Decreasing film quality was associated with decreasing sensitivity, but specificity was not affected by film quality. Because the chest X-ray has suboptimal test characteristics in the diagnosis of pleural thickening and because there is considerable interreader variability, we conclude that caution is required in interpreting radiologic detection of pleural thickening using ILO procedures.