PET is a potentially useful modality for response analysis and prognosis prediction in patients with high-grade non-Hodgkin lymphoma (NHL). The thymidine analog 39-deoxy-39-18 F-fluorothymidine ( 18 F-FLT) was recently introduced as a new tracer. 18 F-FLT uptake correlates with tumor cell proliferation and is suggested to reflect early response to treatment. We performed a prospective study to evaluate the prognostic value of early interim 18 F-FLT PET in patients with NHL. Methods: Patients with untreated NHL were enrolled between 2005 and 2007. Among them, 61 pairs of 18 F-FLT PET images were obtained at baseline (pre), after 1 cycle of chemotherapy (interim), and at the end of all scheduled first-line chemotherapy (final). All 18 F-FLT PET scans were interpreted by quantitative methods (maximum standardized uptake value [SUVmax] and mean standardized uptake value [SUVmean]). Receiver-operating-characteristic curve analysis was performed to define 18 F-FLT PET positivity using a cutoff value predicting disease progression, relapse, or death. Survival outcome was measured by progression-free survival (PFS) and overall survival (OS) rates. Results: Receiver-operating-characteristic curve analysis of SUVmax for prediction of disease progression and death showed the highest area under the curve (AUC) in interim 18 F-FLT PET scans (AUC of 0.841 for PFS and 0.834 for OS, with a cutoff of 1.86; P , 0.001), compared with pre and final 18 F-FLT PET scans. The SUVmean in interim 18 F-FLT PET scans also showed better prediction (AUC of 0.842 for PFS and 0.824 for OS, with a cutoff value of 1.65; P , 0.001) than pre and final 18 F-FLT PET scans. Patients with an interim 18 F-FLT PET SUVmax more than 1.86, who were defined as the interim PET-positive group, were associated with worse 5-y PFS and OS rates than the interim PET-negative group (for PFS: 52.0% vs. 80.7%, respectively, and P , 0.001; for OS: 56.2% vs. 81.4%, respectively, and P , 0.001). By multivariable analysis, the prognostic value of interim 18 F-FLT PET positivity by SUVmax remained significant after adjustment with other prognostic factors (for PFS: hazard ratio, 7.82, 95% confidence interval, 1.65-36.96, and P 5 0.009; for OS: hazard ratio, 5.55, 95% confidence interval, 1.47-33.77, and P 5 0.014). Conclusion: In patients with aggressive NHL, early interim 18 F-FLT PET is a significant predictor of PFS and OS. Early 18 F-FLT PET imaging also has a potential to identify patients with delayed response and nonfavorable prognosis despite achieving a clinical complete response.