FDG-PET Falls Short for Diagnosis of Non–Small Cell Lung Cancer

Get Permission

The ability to make a diagnosis of non–small cell lung cancer (NSCLC) with FDG-PET varies widely among centers, and this imaging modality performed more poorly in community and academic centers than in published studies.1 A secondary analysis of the prospective American College of Surgeons Oncology Group (ACOSOG) Z4031 trial based on written radiology reports found that FDG-PET had lower sensitivity and specificity in patients with known or suspected NSCLC than has been shown in previously published trials. No additional review of FDG-PET images was undertaken for the trial.

“Current [National Comprehensive Cancer Network (NCCN)] guidelines recommend FDG-PET for diagnosis of suspected NSCLC, based on studies showing a high degree of accuracy. We found that FDG-PET performs poorly at single institutions…. The results of PET scans in this population should be interpreted cautiously,” said lead author Eric L. Grogan, MD, Vanderbilt University Medical Center, Nashville.

Previous meta-analyses of FDG-PET used to diagnose NSCLC showed 94% sensitivity and 83% specificity, but the imaging was done at select centers with more experience. ACOSOG Z4031 was conducted at 51 different sites at 39 cities in the United States and found the sensitivity of the scan was 82% and the specificity was 31%.

Study Details

The study evaluated the accuracy of FDG-PET in diagnosing NSCLC in 682 patients enrolled in the trial between 2004 and 2006. All patients had clinical stage I known or suspected NSCLC and underwent surgical resection. The sensitivity, specificity, overall accuracy, and positive and negative predictive value of FDG-PET scans was calculated for all patients as well as for sites with more than 25 participants.

Among 682 study participants (all of whom had preoperative FDG-PET scans), there were 566 cancers and 116 benign lesions. The malignancy rate in patients with FDG-PET scans was 83%; accuracy for diagnosing lung cancer was 73%. The positive predictive value was 85%, and the negative predictive value was 26%. At the eight cities with more than 25 participants, the sensitivity varied significantly—for example, 67% in Los Angeles and 91% in Durham (P =.03).

The accuracy of FDG-PET increased with lesion size. The accuracy was < 50% in lesions < 20 mm, but > 80% in lesions > 30 mm. Accuracy did not improve among subgroups of lesions larger than above 30 mm—that is, accuracy was similar for lesions 31 to 50 mm, 51 to 70 mm, and > 70 mm.

The investigators concluded that more research is needed to determine why performance of the scans varied so widely and did not match up with published studies.

Disclosure: Dr. Grogan reported no potential conflicts of interest.


1. Grogan EL, Deppen SA, Ballman KV, et al: Accuracy of FDG-PET to diagnose lung cancer in the ACOSOG Z4031 trial. 2012 ASCO Annual Meeting. Abstract 7008. Presented June 4, 2012.