The Gordian knot of interim 18-fluorodeoxyglucose positron emission tomography for Hodgkin lymphoma: a meta-analysis and commentary on published studies.

We assessed the diagnostic performance of interim 18-fluorodeoxyglucose positron emission tomography (FDG-PET) with regard to the final outcome of adult patients with newly diagnosed Hodgkin lymphoma (HL). The predefined outcome was treatment failure at the end of follow-up. Bivariate meta-analysis of diagnostic data was used to calculate combined (pooled) estimates. Demographics, quality data and study characteristics were used as potential moderators of outcome in subgroup analysis and meta-regression. A total of 14 studies (16 arms stratified by staging or therapy, 1328 evaluable patients) were deemed eligible for final analysis, after excluding intention-to-treat studies (i.e. those where treatment decision was based on interim PET). The combined effect (95% confidence interval) for sensitivity was 0.67 (0.57-0.76) and specificity 0.89 (0.84-0.93). The corresponding likelihood ratios (LRs) were 6.2 (3.9-10.0) for LR + and 0.37 (0.27-0.50) for LR -, with moderate heterogeneity (I(2) =67%). The estimated negative predictive value was 0.93 (0.85-1.00). The diagnostic performance was influenced by most covariates tested, including age, duration of follow-up, criteria used and time of interim PET. Interim PET retains a high specificity for final outcome, but the sensitivity is low. The use of a PET + study as a surrogate marker is hampered by inconsistent interpretation criteria and study populations. However, the high negative predictive value may permit treatment stratification based on a negative outcome.