EORTC1527/JCOG1609INT/ESSO02: Diffusion-weighted magnetic resonance imaging (DW-MRI) assessment of initially unresectable liver metastasis to improve surgical planning (DREAM)—Primary analysis.

257 Background: Disappearing liver metastases (DLMs) diagnosed on post-chemotherapy (Cx) computed tomography (CT) is a favorable prognostic factor in patients (pts) with colorectal liver metastases (CRLM). However, the optimal treatment of DLMs - whether they should be resected or left behind - is controversial. This is a prospective, multi-centred, international study examining the added value of MRI (DWI, T1/T2 and contrast-enhanced) to that of CT alone in accurate assessment of the viability of DLMs. Methods: Pts with initially unresectable CRLM downstaged to a planned liver resection after Cx were enrolled, based on the obligatory decision of a multidisciplinary team. Pts were imaged by both CT and MRI at baseline and presurgical timepoints. DLMs were defined as disappeared lesions diagnosed by CT alone, while confirmed DLMs (cDLMs) were defined as lesions that disappeared on both CT and MRI. cDLMs were either resected or followed-up for 2 years if not resected. All imaging scans were collected centrally for quality assurance. The primary endpoint was the negative predictive value (NPV) of MRI and CT in confirming the status of cDLMs using either pathological complete response or the absence of a local recurrence at the site of cDLMs during the 2 year follow up. The study was aimed at excluding a NPV ≤0.85 with a 1-sided alfa of 5% and power of 90% under the alternative that the NPV ≥0.95. The planned sample size was 92 evaluable (resected or left behind) cDLMs, assuming a within-patient correlation between cDLMs of 0.2 and an average number of 2 cDLMs per pt. Results: 233 pts were registered at 22 participating centres, and 112 were enrolled for analysis. A median of 8 cycles of Cx was delivered, and pts had a median of 7 CRLMs at baseline. A total of 203 cDLMs were identified while the number of DLMs diagnosed by CT was 296. Of these, 152 cDLMs and 227 DLMs, respectively, were evaluable according to the imaging protocol. Intraoperative ultrasound was performed for 195 cDLMs and, of these, 59 (30.2%) were still detected. The rate of R0/R1 resection was 95.5%. The NPV of evaluable cDLMs either resected or left behind was 62.5 % (95/152, 95% CI:50.8-74.2), which was lower than the prespecified threshold. The NPV of DLMs was 52.9% (119/227, 95% CI:42.7, 63.0). The NPV of resected cDLMs, and those left behind were 56.8% (50/88, 95% CI: 44.2, 69.5) and 70.3% (45/64, 95% CI: 48.6 -92.0), respectively. For DLMs, the NPV of resected DLMs and those left behind were 45.6% (72/158, 95% CI: 35.4-55.7) and 69.6% (48/69, 95% CI: 47.7-91.5), respectively. Conclusions: For pts with initially unresectable CRLM, cDLMs diagnosed by both CT and MRI do not correspond to tumor viability, even after highly effective chemotherapy. Ongoing survival analysis (to be presented at the annual meeting) may impact the treatment strategy of pts with DLMs. Clinical trial information: NCT02781935 .