Improved prognostic value of 18F-FDG PET using a simple visual analysis of tumor characteristics in patients with cervical cancer.

In patients with cervical cancer, it is important to estimate prognosis at the time of diagnosis. This study using PET with (18)F-FDG was undertaken to determine whether a simple and fast visual analysis of characteristics of the primary tumor before initiation of treatment could achieve this goal. METHODS: Forty-seven patients with cervical cancer who were to be treated by combined radiation therapy and chemotherapy were imaged before beginning treatment. They were then followed for up to 3 y for evidence of recurrence or death. Images of the chest, abdomen, and pelvis were obtained 40-90 min after administration of 370-555 MBq (10-15 mCi) (18)F-FDG. Three observers then independently graded the primary tumor for size (0 = small, 1 = moderate, 2 = large), shape (0 = spherical, 1 = nonspherical), heterogeneity of uptake (0 = none, 1 = moderate, 2 = marked), and presence of lymph nodes (0 = none, 1 = pelvic, 2 = paraaortic, 3 = distant). The scores were summed to achieve a total score. A statistical calculation demonstrated that a score cutoff of 4 best separated patients with a good prognosis from patients with a bad prognosis. Kaplan-Meier analysis was used to compute progression-free survival and overall survival. Evaluation of lymph nodes alone was compared with the grading of tumor characteristics. RESULTS: Observers 1 and 2 scored 26 patients as having a good prognosis and 21 as having a bad prognosis. Observer 3 scored 30 and 17, respectively, a statistically insignificant difference. Survival curves were almost identical for the 3 observers. For progression-free survival, approximately 12% of patients with a good score had disease recurrence whereas approximately 75% with a bad score had disease recurrence. For overall survival, approximately 10% (good) and 80% (bad) died. Evaluation of lymph nodes also separated the groups, but not as well as did visual analysis alone. The combination of the 2 was only slightly superior to visual assessment alone. CONCLUSION: A simple, rapid, and highly reproducible system is described for visual grading of characteristics of the primary tumor in patients with cervical cancer at the time of diagnosis. This approach separates patients with a poor prognosis from those who will do well, thus providing a new tool for accurate estimation of prognosis.