Systematic review of response evaluation in gastrointestinal stromal tumors

Abstract

e21500 Background: Our purpose was to conduct an evidence-based systematic review of the literature on the accuracy of 18FDG PET - CT(PET) and contrast enhanced computer tomography (CT) in assessment of response to imatinib therapy in advanced gastrointestinal stromal tumor (GIST) Methods: A systematic review of the literature between January 2000 and December 2008 was performed on Medline and EMBASE. Inclusion criteria included prognostic studies or diagnostic studies using CT or PET to determine response to imatinib. Only articles with Level 1B or higher were included. The following variables were extracted: imaging modality (PET, In-line PET/CT, CT), response criteria used EORTC,RECIST criteria, or novel criteria), and the clinical outcome (time-to-treatment failure (TTF), time to progression (TTP), progression free survival (PFS), disease specific survival (DSS), and/or overall survival). The number of responders and non-responders was extracted from the data, and the reported log rank statistics for the Kaplan Meier curves were used to determine the strength of correlation between the responders and the clinical outcomes. Results: Five articles met the strict criteria for Level 1B evidence. Four articles met the inclusion criteria for In-line PET/CT. In these four articles the log rank correlation between In-line PET/CT response criteria described by the EORTC and the outcome measure of TTF ranged between. p=0.002 -0.04, a very high degree of positive correlation. A subset of four articles met the strict methodological inclusion criteria for CT using RECIST criteria. The log rank correlation for RECIST ranged from p=0.100 - 0.74, no significant correlation. Only one of the articles met the criteria for prospectively applying novel CHOI criteria with CT in a validating cohort. The log rank correlation with the TTF was p=0.0002, a very high degree of positive correlation. Conclusions: In-line PET/CT using EORTC criteria accurately predicts the response to imatinib in terms of TTF or PFS across four Level 1B studies. CT using RECIST criteria has a poor correlation with TTF in well designed studies and is not a useful predictor of response; consistently poor across four Level 1B studies. CT using CHOI criteria shows promise in a single Level 1B study No significant financial relationships to disclose.