Introduction Liver resection prior to resection of the primary cancer is a novel strategy advocated for selected patients presenting with synchronous colorectal liver metastases (sCRLM). This study measured outcomes in patients with sCRLM, following a liver-first (LFA) or classical approach (CA) and used a validated propensity score, the Basingstoke Predictive Index (BPI), to determine differences in survival between case-matched groups.

Method Clinical, pathologic, and complete follow-up (median 34 months) data were prospectively recorded from 582 consecutive patients undergoing hepatic resection for sCRLM at a single centre (2004–2014). 98 patients had a LFA and 467 had a CA to treatment. 17 patients who had a simultaneous bowel and liver resection were excluded. Cumulative disease-free survival (DFS), cancer-specific survival (CSS) and overall survival (OS) were calculated using Kaplan Meier analysis.

Results The LFA group had significantly more patients with rectal cancer (44.9% vs. 27.8%, p = 0.001) or poorly differentiated primaries (17.3% vs. 9%, p = 0.001), had a higher median [interquartile range (IQR)] number of sCRLM [3 (1–20) versus 2 (0–19), p = 0.0007] and size (in millimetres) of sCRLM [17 (3–24), versus 15 (2–29), p = 0.023], compared to the CA patients. Moreover, the median (IQR) preoperative BPI was significantly higher in the LFA group [8.5 (5–10)], versus CA group [8.0 (4–9), p = 0.03]. 73.5% of LFA patients had neoadjuvant chemotherapy, compared to 59.7% in the CA group (p = 0.01). Whilst more LFA patients had a major liver resection (69.4% vs. 56.5%, p = 0.018), there was no difference in Clavien-Dindo grade 3/4 complications (10.2% vs. 7.1%, p = 0.392), median (IQR) length of stay in days [6 (4–9) versus 6 (5–8), p = 0.835] or 30-day mortality [2% vs. 0.2%, p = 0.079] compared to the CA group. The 5 year DFS was 22.8% in the LFA group, significantly lower than the CA group (45.6%, p = 0.001). However there was no difference in 5 year CSS (53.8% vs. 51.1%, p = 0.379) or OS between groups (43.7% vs. 49.6%, p = 0.305). When patients were matched for pre-operative BPI, there was no statistical difference in either 5 year DFS (37% vs. 41.2%, p = 0.083), CSS (53.2% vs. 51.2%, p = 0.616) or OS (44.9% vs. 49.8%, p = 0.846) between groups.

Conclusion In our unit, patients with sCRLM selected for a LFA were a group with more oncologically advanced disease and a poorer prognosis. These patients had a significantly inferior cumulative DFS compared to patients selected for a CA. However, there was no significant difference in cumulative DFS when prognostic variables were matched according to preoperative BPI. These data provide further evidence to support a LFA in selected patients with sCRLM.

Disclosure of interest None Declared.