Background: We showed that non-dimensional criteria (i.e. contrast enhancement and tumor density) correlate with pathologic responses better than RECIST in a group of 38 high-risk STS treated with preoperative CHT +/- RT within a prospective phase III ISG-GEIS trial. We compared the prognostic relevance of RECIST versus CHOI criteria in eligible pts with measurable disease. Methods: From April 2002 to May 2007, 311 eligible pts with localized high-risk STS of extremities or trunk wall received 3 cycles of preoperative epirubicin + ifosfamide +/- RT. Responses by CT or MRI according to RECIST and CHOI were retrospectively evaluated. The prognostic impact of RECIST and CHOI on OS and RFS was evaluated by a univariate Cox regression model. On a dichotomous scale (PR and SD+PD), they were compared in terms of concordance (intechangiability) by Cohen’s k statistics. The pertinent predictive capability was also computed by means of the Harrell c statistic. Results: 242 pts were evaluable by RECIST, 157 by CHOI. Among 114 patients who received CHT alone, 64 were evaluable by both RECIST and CHOI. In these pts, responses by binary CHOI criteria correlated with OS (PD/SD vs PR: HR 4.29, 95%CI 1.84-9.98), with a trend for RFS, while the prognostic impact of responses by binary RECIST was negligible (OS: PD/SD vs PR: HR 0.97, 95%CI 0.38-2.47; RFS: PD/SD vs PR: HR 0.67, 95%CI 0.29-1.57). A reasonable agreement was found between the two diagnostic criteria, with a k statistic value of 0.25 (95%CI: 0.11-0.38). The predictive capability of Choi was greater than RECIST for both OS (0.68 vs 0.50) and RFS (0.58 vs 0.54). In all pts (242 by RECIST and 157 by Choi), we observed a better outcome for both OS and RFS in pts achieving a PR, as by RECIST (for OS: PD vs PR: HR 2.28, 95%CI 1.16-4.46; for RFS: PD vs PR: HR 2.11, 95%CI 1.12 - 3.98) and CHOI (for OS: PD vs PR: HR 5.7, 95%CI 2.81-11.64; for RFS: PD vs PR: HR 3.09, 95%CI 1.39 – 6.89). Conclusions: Response to preoperative CHT+RT were associated to a better outcome. CHOI criteria were better predictors than RECIST in pts receiving preoperative CHT alone.