Although studies have shown that 18 F-FDG PET, when used to assess the response of malignant lymphoma after treatment, has a strong ability to predict relapse, its diagnostic accuracy in clinical practice remains unclear. The aim of this study was to systematically review the diagnostic accuracy of 18 F-FDG PET in detecting residual disease at the completion of first-line therapy of Hodgkin's disease (HD) and aggressive non-Hodgkin's lymphoma (NHL). Methods: We searched relevant articles from 1966 to July 2006 using MEDLINE, EMBASE, SCOPUS, Biological Abstracts, bibliographies, review articles, and textbooks without language restriction. One assessor (for non-English-language studies) or 2 assessors (for English-language studies) independently reviewed each article to abstract relevant study characteristics and results. Relevant individual patient data or subgroup data were provided by the investigators if they were unavailable from the publications. We estimated summary receiver operating characteristic curves and confidence regions for summary sensitivity and specificity. Results: Nineteen studies consisting of 474 HD and 254 aggressive NHL patients were included. These studies had heterogeneity and suboptimal methodologic quality and reporting. Reported ranges for the sensitivity and specificity of 18 F-FDG PET in predicting disease relapse were 0.50-1.00 and 0.67-1.00, respectively, for HD and 0.33-0.77 and 0.82-1.00, respectively, for NHL. These estimates were similar when conventional imaging tests showed a residual mass. For HD studies, the summary receiver operating characteristic curves were similar irrespective of whether a residual mass was detected by conventional tests. Factors explaining the variability of diagnostic estimates were not identified. Conclusion: Although currently available evidence is still limited, 18 F-FDG PET seems to have good diagnostic accuracy for assessing residual HD at the completion of first-line treatment. Clinical data on this use of 18 F-FDG PET for aggressive NHL are more limited. Prospective studies with a more rigorous research design, conduct, and reporting would more reliably reveal the clinical diagnostic accuracy of this imaging modality.