Accuracy of fluorodeoxyglucose-positron emission tomography within the clinical practice of the American College of Surgeons Oncology Group Z4031 trial to diagnose clinical stage I non-small cell lung cancer

Abstract

Background: Fluorodeoxyglucose-positron emission tomography (FDG-PET) is recommended for diagnosis and staging of non-small cell lung cancer (NSCLC). Meta-analyses of FDG-PET diagnostic accuracy demonstrated sensitivity of 96% and specificity of 78% but were performed in select centers, introducing potential bias. This study evaluates the accuracy of FDG-PET to diagnose NSCLC and examines differences across enrolling sites in the national American College of Surgeons Oncology Group (ACOSOG) Z4031 trial.

Methods: Between 2004 and 2006, 959 eligible patients with clinical stage I (cT1-2 N0 M0) known or suspected NSCLC were enrolled in the Z4031 trial, and with a baseline FDG-PET available for 682. Final diagnosis was determined by pathologic examination. FDG-PET avidity was categorized into avid or not avid by radiologist description or reported maximum standard uptake value. FDG-PET diagnostic accuracy was calculated for the entire cohort. Accuracy differences based on preoperative size and by enrolling site were examined.

Results: Preoperative FDG-PET results were available for 682 participants enrolled at 51 sites in 39 cities. Lung cancer prevalence was 83%. FDG-PET sensitivity was 82% (95% confidence interval, 79 to 85) and specificity was 31% (95% confidence interval, 23% to 40%). Positive and negative predictive values were 85% and 26%, respectively. Accuracy improved with lesion size. Of 80 false-positive scans, 69% were granulomas. False-negative scans occurred in 101 patients, with adenocarcinoma being the most frequent (64%), and 11 were 10 mm or less. The sensitivity varied from 68% to 91% (p=0.03), and the specificity ranged from 15% to 44% (p=0.72) across cities with more than 25 participants.

Conclusions: In a national surgical population with clinical stage I NSCLC, FDG-PET to diagnose lung cancer performed poorly compared with published studies.

Figure 1

Enrolling site location with size of circle corresponding to participation volume – 51 sites in 39 cities. Individual dots are participants by zip code at time of enrollment. Dots are overlapping for those with identical zip codes.

Figure 2

FDG-PET accuracy improved with lesion size from 67% in lesions that were 1–2cm (sensitivity 76% and specificity 35%) to 84% in lesions that were 3 to 5 cm (sensitivity 90% and specificity 18%) (p<0.001 using an analysis of variance test). Accuracy of FDG-PET to diagnose lung cancer by lesion size in millimeters. Accuracy = (True Positives + True Negatives)/Total Population