Interobserver variability and accuracy of computed tomographic assessment of nodal status in lung cancer

Ann Thorac Surg. 1994 Jul;58(1):158-62. doi: 10.1016/0003-4975(94)91091-x.


To assess the interobserver variability of computed tomography in determining nodal status in non-small cell lung carcinoma, four experienced radiologists reviewed the computed tomographic scans of 147 patients. Interobserver variability was calculated using the kappa statistic. In addition, the accuracy of CT assessment of the nodal status by the four observers was measured by comparing their findings with thorough mediastinal exploration at both mediastinoscopy (n = 35) and thoracotomy (n = 112). Interobserver variability was large between the four radiologists regarding nodal status on a per-patient basis (kappa = 0.38). Sensitivity of computed tomography for the observers on a per-patient basis ranged from 40% to 69% with a 1.0-cm criterion and from 28% to 56% with a 1.5-cm criterion. From the large interobserver variability and the low sensitivities in this study it can be concluded that a negative result of computed tomography regarding mediastinal lymph nodes does not eliminate the need for mediastinoscopy or exploration of the mediastinum at the time of operation in patients with non-small cell lung cancer.

Publication types

  • Comparative Study

MeSH terms

  • Carcinoma, Non-Small-Cell Lung / diagnostic imaging
  • Carcinoma, Non-Small-Cell Lung / epidemiology
  • Carcinoma, Non-Small-Cell Lung / secondary*
  • Humans
  • Lung Neoplasms / pathology*
  • Lymph Nodes / diagnostic imaging
  • Lymphatic Metastasis
  • Mediastinoscopy
  • Observer Variation
  • Sensitivity and Specificity
  • Thoracotomy
  • Tomography, X-Ray Computed / statistics & numerical data*