Real-world outcomes and recurrence risk factors of liver transplantation for permanently unresectable colorectal liver metastases following the TransMet trial.

3589 Background: Following the TRANSMET trial that demonstrated an overall survival (OS) benefit of liver transplantation (LT) over chemotherapy (CT) alone for selected patients with permanently unresectable colorectal liver metastases (uCRLM), the program of LT has continued in France with the same eligibility criteria, independent validation and graft allocation policy. We assessed the real-world oncological outcomes and prognostic factors for post-LT recurrence-free survival (RFS) in the whole cohort. Methods: This nation-wide analysis pooled both patients transplanted in TransMet (Trial cohort) and those after trial closure (Extension cohort). OS and RFS were evaluated in each cohort. Multivariable Cox models for RFS using sensitivity analyses including forced cohort adjustment and bootstrap resampling were performed. Results: A total of 87 patients underwent LT (37 Trial cohort; 50 Extension cohort). Baseline characteristics at CRLM diagnosis were comparable. At LT, patients in the Extension cohort received significantly fewer CT cycles (≥24 cycles, 16% vs 41%, p=0.01) and showed higher rate of partial response (94% vs 57%, p < 0.001) after first-line CT. At time of LT, tumor burden, RECIST status and tumor markers were similar. Median follow-up was shorter in the Extension cohort (16 vs 57 months). No significant differences in OS and RFS were observed between extension and trial cohorts (30-months OS 89% vs 81%; 30-months RFS 47% vs 39%, respectively). In univariable analysis, risk factors of RFS included ≥2 CT lines, ≥24 CT cycles before LT, CEA at LT > 5 ng/mL, progression on CT and interval from primary tumor surgery >2 years. In multivariable Cox analysis, ≥2 lines or >24 cycles before LT (HR 2.90, 95% CI 1.32–6.36; p<0.01) and CEA at LT >5 ng/mL (HR 2.26, 95% CI 1.18–4.33; p=0.01) remained independently associated with RFS, adjusted on interval from primary tumor surgery. Bootstrap resampling (2,000 iterations) highlighted these results in 82% and 67% of models, respectively. Conclusions: LT for uCRLM gives good and reproducible oncological outcomes across trial and extension cohorts. This exploratory analysis supports that RFS is impaired by a prolonged pre-transplant chemotherapy (≥ 2 lines or ≥ 24 cycles) and improved by a good biological response (normalization of CEA), supporting considering LT in a multidisciplinary setting as soon as definitive unresectability is established.