Abstract Background To assess the additional value of density measurement using contrast-enhancement sequences (Choi assessment) in a real-life cohort of adult soft tissue sarcoma patients treated with trabectedin. Methods Eligibility criteria included adults (age ⩾18) treated between 01/2007 and 12/2011, with at least two trabectedin cycles after failure or intolerance to doxorubicin/ifosfamide. Baseline and first computed tomography (CT)-scans were centrally reviewed by an experienced radiologist. Results The retrospective cohort consists of 134 (73 female) patients treated with trabectedin 1.5 mg/m2 given as a 24-h infusion every 3 weeks. Patients received a median of five trabectedin cycles (range: 2–33) and the main cause of discontinuation was progressive disease (PD) ( n = 105, 78.4%). Response Evaluation Criteria in Solid Tumours (RECIST) assessment was feasible in 128 (95.5%) patients, with Choi assessment performed in 92 (68.7%) patients, generally due to inadequate sequences or exclusive lung metastases. Concordance between both methods was fair (Kappa = 0.290). We identified five patients with false PD (i.e. PD according to RECIST but stable disease/partial response as per Choi). Univariate analysis did not identify any predictive factors for false PD. Median overall survival (OS) of patients with PD as per RECIST but stable disease/partial response (SD/PR) according to Choi was better than for patients with PD according to both RECIST and Choi (14 months versus 8 months; p = 0.052). Conclusions Choi assessment may identify patients with false PD who achieved improved efficacy outcomes, suggesting that trabectedin may delay tumour progression even in the case of non-dimensional response. Dual size and tumour density assessment may be more suitable to evaluate responses to trabectedin in sarcoma patients as well as to improve the decision-making strategies for the continuation of trabectedin therapy.