Surgery for colorectal liver metastases: the impact of resection margins on recurrence and overall survival

The selection criteria for surgery in our centre included a sufficient remaining tumour-free liver volume (30%) with adequate blood perfusion and bile drainage, and absence of: a) non-resectable extrahepatic metastases, and/or b) no disseminated disease as evaluated preoperatively. Patients with extrahepatic disease and R2 resections were excluded from the current study. Preoperative investigations included computed tomography (CT) scan of the chest and abdomen/pelvis, and tumour marker analysis (CEA: carcinoembryonal antigen). In cases with an inconclusive CT scan, magnetic resonance imaging (MRI) of the liver, contrast-enhanced ultrasound and 18 F-fluorodeoxyglucose ¹⁸(FDG)-positron emission tomography (PET)/CT scan were performed. Each patient was discussed in a multidisciplinary team meeting with surgeons, oncologists and radiologists.

Preoperative chemotherapy (n = 88) was given in a perioperative setting (n = 43) or as a downstaging procedure (n = 40) in patients with initially deemed unresectable disease. Five patients developed CLM during adjuvant treatment with chemotherapy after resection of stage III colon cancer. We evaluated the outcome of chemotherapy by the Response Evaluation Criteria in Solid Tumour (RECIST) version 1.1 [17]. The size of the metastases was measured on CT scan by dedicated radiologists. All patients in the perioperative group were offered the FOLFOX regimen (fluorouracil, leucovorin and oxaliplatin) with an intended six cycles before and after surgery. They were evaluated with CT scan after three and six cycles. The indication for perioperative chemotherapy has changed during the period. A total of 17 patients were enrolled in the European Organisation for Research and Treatment of Cancer (EORTC) multicentre study 40983 and randomised for surgery alone (n = 7), or surgery with perioperative chemotherapy (n = 10) in the period 2001 to 2004 [18]. After that, patients <76 years with Eastern Cooperative Oncology Group (ECOG) performance status 0 to 1 and no previous treatment with oxaliplatin had been offered perioperative chemotherapy. In the downstaging group, patients were treated with several different chemotherapy regimens as listed in Table 1. First-line treatment with the Nordic FLOX or FLIRI regimen was most commonly used, optionally in combination with EGFR (endothelial growth factor receptor) inhibitors or angiogenesis inhibitors.

Surgical techniques included subcostal incision, intraoperative ultrasonography, occasionally repeated inflow control (the Pringle manoeuvre), and transection using Ultracision, Kelly clamp and Cavitron Ultrasonic Surgical Aspirator (CUSA). Throughout the period we have intended to achieve a parenchyma-sparing approach, with wedge resections whenever possible. Formal resections (hemihepatectomies or lobectomies) have been reserved for metastases placed centrally or near the hepatic veins. To increase intended complete tumour eradication, intraoperative radiofrequency ablation, and portal vein ligations/embolization with two-stage resections have been performed. Simultaneous colorectal cancer surgery has been reserved for healthy patients with colon cancer and less advanced CLM. Further details are listed in Table 1.

Follow-up after surgery included CT scan of the chest, abdomen and pelvis every three months for the first two years, and thereafter every six months for the next three years. Serum level of CEA tumour marker was obtained every third month. We defined LR by CT scan as a new lesion in contact with the previous resection surface. The resected area was easily detected with CT due to the wide use of metallic clips during the transection. The data were based on the first detection of recurrence. Data of recurrence were not available in four patients. Patterns of recurrence were stratified according to LR, hepatic recurrence (without LR) and extrahepatic recurrence. LR included patients with a) LR only, b) LR and relapse in other sites of the liver and c) LR with concomitant new extrahepatic lesions. During the follow-up, thirteen patients died from causes other than colorectal cancer, and six from treatment-related causes. These patients were also included in the analysis of OS, according to the definition stated by Punt et al.[19].

Variables with possible impact on OS like RM, age, size, number of metastases, bilobar distribution, disease-free interval and TNM stage of primary tumour were analysed with univariate and multivariate survival methods. The exact chi-square (χ²) test was used for categorical variables, the t test for normally distributed variables, and the Mann–Whitney U (MWU) test for non-normally distributed continuous variables. The Kruskal-Wallis one-way analysis of variance test was used to compare more than two non-normally distributed samples. Multinomial logistic regression was used to evaluate LR in relation to RMs. Survival was estimated by the Kaplan-Meier method [20] and tested for significance with the log-rank test [21]. Multivariate analysis was performed as Cox proportional regression [22]. Continuous predictors such as RMs were also modelled using multiple fractional polynomial regressions [23]. A P value ≤0.05 was considered significant. OS was defined as time from resection to death irrespective of cause, and TTR was defined as the interval between resection and the detection of a local or distant relapse [19]. All analyses were performed using SPSS Statistics version 19 (IBM Corp., Armonk, NY, USA) and Stata 12 statistical software (StataCorp, College Station, TX, USA). We decided to use TTR rather than disease-free survival as a parameter in assessing recurrence patterns, since the latter has treatment-related and non-cancer-related deaths as endpoints, which could be misleading according to the definition by Punt et al.[19].

The Regional Committee of Ethics of Western Norway Health Authority approved the study, with an exemption to the requirement for obtaining informed consent from patients included in the retrospective part (1998 to 2008). In the prospective part (2008 to 2010) patients were enrolled through written consent.

R1 patients had more advanced disease compared to R0 according to bilobar locations (P = 0.007, χ²-test) and number of metastases (P = 0.099, MWU test). There was no significant difference between R0 and R1 patients in the TNM status of primary tumour in colon or rectum, American Society of Anesthesiologists (ASA) score, size of the metastases and the use of preoperative chemotherapy. Postoperative chemotherapy was administered more frequently in R1 patients (n = 12 of 48, 25.0%) compared to R0 (n = 31 of 205, 15.1%), P = 0.016, χ²-test. In the chemotherapy group, there was no difference in number of R1 resections between patients with partial response or stable disease using the RECIST criteria (P = 0.575, χ²-test). In the perioperative and the downstaging group a total of thirty (69.8%) and seven (17.5%) patients, respectively, underwent postoperative chemotherapy (P <0.0001, χ²-test).

Global recurrent disease occurred in n = 175 (72.3%) patients, whereas involvement of the resection surface was found in 40 cases (16.5%). Further details are listed in Table 2. We found a lower global recurrence in the groups C and D compared to A and B. The risk for recurrence according to RMs (A to D) was assessed with a multinomial logistic regression, as detailed in Table 3. The odds ratios for LR were significantly higher in groups A and B relative to group D. RMs did not seem to impact hepatic recurrence, whereas extrahepatic recurrence was more frequent compared to no recurrence with RMs <5 mm (0.005). A total of 21.5% of the patients were recurrence-free after five years. TTR increased significantly (P = 0.009) with the increasing extent of the RMs (Figure 1a), but this difference was repealed when we omitted those patients (n = 40) with all kinds of LR (P = 0.097). We also detected a non-significant difference in five-year TTR between R0 (24.5%) and R1 (0%, P = 0.127). No additional benefit for TTR was seen with RMs beyond 10 mm, where the groups C and D were nearly equal in outcome (Figure 1a).

A total of 50 of 253 (19.8%) patients underwent a second operation for resectable recurrence. Twenty (40.0%) of these were due to LR after the first resection. In 27 cases (54.0%), recurrence was found in other sites of the liver whereas only three patients (6.0%) had combined intra- and extrahepatic relapses. In patients with LR only, 11 of 14 (78.6%) patients were resected. Of the 48 patients having a primary R1 resection, 15 (31.3%) underwent a second operative procedure.

Five- and ten-year OS survival rates were 38.7% and 23.0%, respectively, whereas median OS was 45.0 months. Five-year OS of R0 vs. R1 was 42.5% and 16.1%, (P = 0.011, Figure 1b), whereas median OS in R0 and R1 were 48.1 and 32.4 months, respectively. By sub-grouping according to margin width (A to D), an increased OS was seen in the univariate analysis (P = 0.035, see Figure 1b). However, there was no extra benefit when the RMs exceeded 10 mm (group C vs. D). Patients were also stratified according to preoperative chemotherapy (n = 88), and the difference in five-year OS between R0 and R1 was maintained (P = 0.037). In the perioperative group (n = 43), a non-significant difference (P = 0.502) in five-year OS was seen between R1 (34.3%) and R0 (54.2%). In the downstaging group (n = 40), the five-year OS was 40.2% for the R0 cases vs. none survivors in the R1 group (P = 0.017). Patients with initially unresectable metastases had more extensive disease evaluated as the average number of metastases (3.6) compared to the perioperative (2.5) and the surgery alone group (2.3), using Kruskal-Wallis test (P = 0.002). Positive RMs also predicted a borderline significant adverse outcome in the Cox proportional hazards model (P = 0.067), along with age, ASA score, number of metastases, size of the metastases and repeated resections (Table 4). When using the RM sub-groups (A to D) in the Cox model (P = 0.111) and the RMs as a continuous variable (P = 0.099), significance was not reached. We neither found any substantial differences in OS with a cut-off margin of 5 mm (P = 0.194). We also applied multiple fractional polynomials in the Cox regression model without identifying any non-linear relationships between RMs and OS. When adjusting for patients offered preoperative chemotherapy, multivariate analyses revealed RMs still to be a borderline significant factor predicting adverse OS (P = 0.081). However, in contrast to the rest, there was an adverse effect on OS of R1 vs. R0 in the downstaging group (test of interaction P = 0.020), adjusting for the same variables as listed in Table 4. No such effect was evident in the perioperative group.

Finally, we conducted survival calculations according to the site of recurrence and the involvement of LR independently of RMs. We could not reveal any difference in five-year OS between LR (total), hepatic-only and extrahepatic recurrence (P = 0.947). Patients with LR only proved a better five-year OS compared to patients with recurrence at other sites (35.9% vs. 25.4%, P = 0.048). Within the latter group we neither found any substantial differences in OS (P = 0.130).