The accuracy of integrated PET-CT compared with dedicated PET alone for the staging of patients with nonsmall cell lung cancer.

BACKGROUND

The treatment of patients with nonsmall cell lung cancer (NSCLC) is determined by the stage. We evaluated the accuracy of staging using integrated positron emission tomography (PET) and computed tomography (CT) and compared it with dedicated PET visually correlated with CT scan.

METHODS

A prospective blinded trial was performed on a consecutive series of patients with NSCLC. Patients underwent integrated PET-CT scanning with 2-[18F]-fluoro-2-deoxy-D-glucose (FDG-18). A radiologist assigned the T, N and M status. No sooner than 2 weeks the same radiologist read the dedicated PET alone, without the integrated CT images and a T, N and M status was assigned again. The most recent CT scan was available and visually correlated with both studies. All patients underwent biopsies of suspicious N2 or N3 lymph node or distant metastases and if negative, pulmonary resection with lymphadenectomy was performed.

RESULTS

There were 129 patients. Integrated PET-CT is a better predictor than PET for all stages of cancer and achieved statistical significance for stage I (52% versus 33%, p = 0.03) and for stage II (70% versus 36%, p = 0.04). It also is a better overall predictor for T status (70% versus 47%, p = 0.001) and the N status (78% versus 56%, p = 0.008). Nodal analysis shows that integrated PET-CT was more accurate for the total N2 nodes (96% versus 93%, p = 0.01) and for the total N1 nodes (90% versus 80%, p = 0.001). It was also more sensitive, specific, and had a higher positive predictive value for both N2 and N1 nodes (p < 0.05 for all). Integrated PET-CT is significantly more sensitive at the 4R, 5, 7, 10 L and 11 stations and more accurate at the 7 and 11 lymph nodes stations than dedicated PET.

CONCLUSIONS

Integrated PET-CT using FDG-18 better predicts stage I and II disease as well as the T and N status of patients with NSCLC when compared with dedicated PET alone. It is more accurate at some nodal stations but still only achieves an accuracy of 96% and 90% for the N2 and N1 nodes, respectively.