This study was performed to investigate the utility of FDG-PET for: (1) initial staging, and (2) restaging of the primary and mediastinal nodal lesions 2 weeks after the completion of preoperative chemoradiotherapy in patients with stage III non-small cell lung cancer (NSCLC). Twenty-six patients with histologically confirmed stage III NSCLC were accrued to this study from April 1993 to July 1998. They included 21 with stage IIIA (N2) NSCLC who were enrolled into an institutional phase II study, and 5 patients with a highly selected subset of stage IIIB disease characterized by the presence of microscopic metastatic disease in contralateral mediastinal lymph nodes who were also treated with preoperative chemoradiotherapy; N3 lesions (n=3) and minimal T4 lesions. Demographic characteristics included median age 62 years (a range from 47 to 73) and gender ratio of male 15 to female 11. Histologic types of tumor consisted of squamous cell carcinoma 6, adenocarcinoma 11, large cell carcinoma 5, and non-small cell carcinoma 4. All patients had FDG-PET imaging of the chest before the initiation and 2 weeks after completion of preoperative therapy. The FDG-PET images were evaluated qualitatively for uptake at the primary tumor sites and mediastinal lymph nodes. Standard uptake values (SUVs) were also calculated for the primary tumors and all PET findings were correlated with surgical histopathologic data. Preoperative chemoradiotherapy resulted in complete pathologic response in 8 of 26 primary lesions. By qualitative analysis, 96% of these tumors showed level 3 or 4 uptake before preoperative chemoradiotherapy. After chemoradiotherapy, 57% (15/26) of patients showed at least a one level decrease in uptake, and the sensitivity and specificity of FDG-PET for differentiating residual tumor from pathologic complete response were 67% (12/18) and 63% (5/8). Mean SUV was 14.87+/-7.11 at baseline and decreased to 5.72+/-3.35 after chemoradiotherapy (n=21, P<0.00001). When a value of 3.0 was used as the SUV cut-off, sensitivity and specificity were 88 and 67%, respectively. The mean values of visual intensity were 3.87+/-0.35 and 3.8+/-0.51 for patients who achieved pathologic complete response (n=8) and for those who showed residual cancer after the preoperative therapy (n=18), respectively. The mean SUVs were 16.97+/-8.52 and 14.03+/-6.61 for patients who achieved pathologic complete response (n=6) and for those who showed residual cancer (n=15) after the preoperative therapy, respectively. Therefore, the degree of FDG uptake before preoperative chemoradiotherapy did not provide predictive value for subsequent tumor response. For mediastinal initial staging, the sensitivity and specificity of FDG-PET were 75 and 90.5%. The sensitivity and specificity of FDG-PET for mediastinal restaging were 58.0 and 93.0%. These results indicate that FDG-PET is useful for monitoring the therapeutic effect of neoadjuvant chemoradiotherapy in patients with stage III NSCLC. For the primary lesions, SUV based analysis has high sensitivity but limited specificity for detecting residual tumor. In contrast, for restaging of mediastinal lymph nodes, FDG-PET is highly specific, but has limited sensitivity.