Hrudaya Nath

MD

University of Alabama at Birmingham

Radiologist

NLST (Learn more about this study)

2006-00391

Feb 20, 2006

Observer Variability among PLCO-NLST Radiologists in Interpretation of T-0 CXR

Interpretation of chest radiographs in the NLST requires detection of abnormalities, and their classification into categories defined by the study protocol. The main distinctions made by the readers are 1) whether a lung nodule is present, 2) whether detected nodules contain a benign pattern of calcification, and 3) other abnormalities such as hilar / mediastinal adenopathy, lobar atelectasis, chest wall mass, and / or pleural effusion, which may also indicate the presence of lung cancer. It is long known that error rate for chest radiographic detection of lung cancer is considerable, ranging from 20% to 50%. Visual perception of lung nodules is dependent upon its conspicuity, and that in turn depends upon many variables including technical factors such as screen / film combination, exposure and processing factors, as well as the size, density, and location of the nodule. The digital radiographic techniques and soft copy viewing have achieved greater uniformity in the appearance of the radiographic image, as well as modest improvement in the recognition of lung nodules, particularly in the retrocardiac lung. While there is considerable literature about the sensitivity / specificity and limitations of detecting lung nodules by chest radiography, there are limited data about reader variability in lung cancer screening, particularly since the introduction of digital radiographic techniques. Therefore, this study is undertaken to document the level of agreement in screening CXR interpretation among the PLCO-NLT radiologists, similar to the first CT reader variability exercise.