Clinical Application of Positron Emission Tomography (PET) in Designing Radiation Fields in Non-Small Cell Lung Cancer (NSCLC) Patients

Reviewer: Christopher Dolinsky, MD
University of Pennsylvania School of Medicine
Last Modified: October 19, 2005

Presenter: Charles G. Wood, MD
Presenter's Affiliation: University of Pennsylvania, Dept. of Radiation Oncology, Philadelphia, PA
Type of Session: Scientific


  • PET scans offer quantitative data regarding the glucose utilization of both tumor and normal tissue.
  • PET scans have become commonplace in the staging of non small cell lung cancer (NSCLC)
  • Studies have looked at the accuracy of PET scanning for these patients in the past, but there is little information about the accuracy of PET scans by particular tumor and lymph node locations; for example the primary tumor, the mediastinum and the hilar regions.
  • There is interest in incorporating information from PET scanning into radiation treatment planning, and this cannot be accomplished without better information about PET accuracy.

Materials and Methods

  • This is a retrospective study that looked at 351 patients with NSCLC at the University of Pennsylvania who underwent preoperative PET scans between 1/03 and 7/05.
  • Of this population, 257 patients did not undergo any preoperative therapy and were included in this analysis.
  • PET findings regarding the primary site, the ipsilateral hilum and the mediastinum were recorded and correlated with surgical pathology at the time of resection.
  • The analysis was done based on the radiology reports from the PET center, not by examining the actual PET images.
  • Accuracy, sensitivity, specificity, positive predictive value and negative predictive value were all determined for PET imaging.
  • PET results were scored as positive, suspicious, or negative based on the radiology report.


  • 70% of PETs were done in an academic setting and 30% were in a community setting.
  • The median age of patients was 68 years (range 35-86) and the majority of patients were early stage (46% Stage IA, 17% Stage IB).
  • The sensitivity and accuracy for the primary tumor were both 95%, and the specificity for the primary tumor was not computed.
  • The sensitivity, specificity, and accuracy for the mediastinum were 49%, 95%, and 84%.
  • The sensitivity, specificity, and accuracy for the ipsilateral hilum were 50%, 84% and 80%.
  • The positive predictive value and negative predictive value for the mediastinum were 70% and 86%.
  • The positive predictive value and negative predictive value for the ipsilateral hilum were 31% and 92%.
  • There was a decreased accuracy for PET seen for bronchoalveolar primary tumors compared to other histologies (86% vs 96%, p=0.02)
  • There was decreased accuracy of PET for detecting primary tumors <1cm compared to tumors >1cm (78% vs 96%, p=0.01)
  • PET stage and pathologic stage were the same in 51% of cases, PET stage was higher in 14% of cases, and PET stage was lower in 35% of cases.

Author's Conclusions

  • PET imaging is a potentially important tool for designing radiation treatment fields for lung cancer.
  • PET needs to be used with caution when considering nodal regions because the positive predictive value is low.
  • The negative predictive value is high for PET, and may help exclude large treatment volumes in selected cases.
  • Further evaluation of PET imaging site by site in more advanced stage patients is warranted.

Clinical/Scientific Implications

The authors should be commended for asking an important question that has a major impact on clinical decision making. PET imaging has become quite popular, and many clinicians are using PET findings to make significant treatment decisions about the sizing of radiation fields without a lot of evidence regarding the accuracy of PET. Dr. Wood presented a sobering look at the accuracy of PET – especially for its low positive predictive value when it comes to the mediastinal and hilar regions. Dr. Wood's research shows a somewhat lower accuracy of PET than has been previously published. One explanation for this is that this series is larger and perhaps more heterogeneous than previous single institution series. The advent of PET/CT scans may improve the accuracy of metabolic imaging, and this has been suggested in a randomized trial. However, particularly in patients who have small volumes of disease (<1cm) or questionable mediastinal disease that will greatly alter the radiation plan, it may be prudent to send them for mediastinal node sampling. Mediastinal lymph node sampling is the current gold standard for accurate lung cancer staging.