Determinants of Long-Term Outcome in Patients Undergoing Simultaneous Resection of Synchronous Colorectal Liver Metastases

Background It remains unclear which patients can benefit from simultaneous resection of synchronous colorectal liver metastases (SCRLMs). This study aimed to examine the prognostic value of patient- and tumor-related factors in predicting long-term outcomes of patients undergoing simultaneous resection of SCRLMs and to help patients select a suitable therapeutic regimen and proper surveillance. Methods Clinicopathological and outcome data of 154 consecutive SCRLM patients who underwent simultaneous resection between July 2003 and July 2013 were collected from our prospectively established SCRLM data and analyzed with univariate and multivariate methods, and the prognostic index (PI) was formulated based on the regression coefficients (β) of the Cox model. The patients were classified into high- and low-risk groups according to the PI value; the cut-off point was the third quartile. Results The 5-year overall survival rate was 46%, and the 5-year disease-free survival rate was 35%. Five factors were found to be independent predictors of poor overall survival (OS) by multivariate analysis: positive lymph node status, vascular invasion, BRAF mutation, the distribution of bilobar liver metastases (LMs) and non-R0 resection of LMs. Compared to low PI (≤5.978), high PI (>5.978) was highly predictive of shorter OS. Three factors were found to be independent predictors of poor disease-free survival (DFS) by multivariate analysis: tumor deposits, BRAF mutation and bilobar LM distribution. We also determined the PI for DFS. Compared to low PI (≤2.945), high PI (>2.945) was highly predictive of shorter DFS. Conclusions Simultaneous resection of SCRLM may lead to various long-term outcomes. Patients with low PI have longer OS and DFS, while those with high PI have shorter OS and DFS. Thus, patients with high PI may receive more aggressive treatment and intensive surveillance, This model needs further validation.


Introduction
Colorectal cancer (CRC) is the third most commonly diagnosed cancer in males and the second in females, with over 1.2 million new cancer cases and 608,700 deaths estimated to have occurred worldwide in 2008 [1]. Approximately 25% of CRC patients present with synchronous colorectal liver metastases (SCRLMs) at the time of initial diagnosis, and 25,50% of patients subsequently develop metachronous colorectal liver metastases (MCRLMs) at the time of disease recurrence [2]. The median survival is 2.3 to 21.3 months for patients with unresectable liver metastases (LMs) [3]. A subset of patients with either SCRLM or MCRLM benefit from radical surgery and may even achieve a cure; approximately 25,50% of patients with surgically resected colorectal liver metastasis (CRLM) survive 5 or more years [4][5][6][7].
Unfortunately, the optimal surgical strategy for patients presenting with SCRLM remains controversial. Traditionally, patients have undergone a two-stage procedure with resection of the primary colorectal tumor followed by chemotherapy and subsequent liver resection. Presently, an increasing number of studies have demonstrated that simultaneous resection is an acceptable and safe option due to advances in the surgical technique of liver resection and enhancements in anesthesia and critical care [3,4,8,9]. This strategy has potential benefits in terms of cost and quality of life, particularly as using contemporary chemotherapy regimens may damage the liver parenchyma and leave patients ineligible for surgical resection [3].
Several previous studies have reported prognostic factors for patients undergoing two-stage resection [5,[8][9][10]; however, few have examined the factors that influence long-term outcome in patients exclusively undergoing simultaneous resection, and the sample size was very small [4,11].
This study aimed to examine the prognostic value of patientand tumor-related factors in predicting long-term outcomes in a cohort of prospectively collected patients undergoing simultaneous resection of SCRLMs and to help patients to select a suitable therapeutic regimen and proper surveillance.

Inclusion Criteria
We reviewed our prospectively collected SCRLM database between July 2003 and July 2013 and identified 154 consecutive patients treated with simultaneous curative surgery for initially resectable SCRLM. The selection criteria for simultaneous surgery have been reported previously: [12] expected radical resection of primary disease and margin-negative resection of liver LMs (R0), no unresectable extrahepatic disease and adequate predicted volume of the post-resection hepatic remnant. All deaths occurred were due to colorectal cancer. Patients who underwent previous hepatic resections or ablations of the CRLM, experienced perioperative death or had incomplete materials were excluded from this study. The study was approved by the institutional review board of Zhongshan Hospital. All of the patients provided written consent.

Data Collection
Clinicopathological data for each patient were collected from our prospectively collected SCRLM database. The tumors were staged using the 7th edition of The International Union Against Cancer (UICC)/American Joint Committee on Cancer (AJCC) Tumor, Node, Metastasis (TNM) classification system [13]. KRAS and BRAF mutations were detected by pyrosequencing, and the timing of perioperative chemotherapy was recorded. The followup regimen included routine computed tomography of the chest, abdomen and pelvis and regular colonoscopic surveillance. Disease recurrence was recorded on the basis of clinical, endoscopic or radiological findings at the time of diagnosis. The date of last follow-up, vital status and recurrence of LM were determined and recorded for all patients. Overall survival (OS) was calculated from the date of diagnosis to the date of death due to CRC or last follow-up. Disease-free survival (DFS) was measured from the date of surgery until the date of documented disease recurrence.

Statistical Analysis
Summary statistics were obtained using established methods and are presented as percentages or median values with standard deviations. Categorical data are summarized as percentages and were analyzed using chi-squared analysis or Fisher's exact test. OS and DFS were analyzed using the Kaplan-Meier method; survival curves were compared using the log-rank test. Univariate and multivariate analyses were performed using the Cox proportional hazards model; significant prognostic factors in univariate analysis were entered into the Cox proportional hazards model using stepwise selection to identify independent predictors. The prognostic index (PI) of patients was calculated based on the regression coefficients (b) of the Cox model: PI = b161+b262 +…+ bn6n. The patients were classified into high-and low-risk groups according to the PI value; the cut-off point was the third quartile [14][15][16]. All of the statistical analyses were performed using SPSS 16.0 software (SPSS, Chicago, IL, USA). Two-sided P-values were calculated, and P,0.05 was considered significant.

Clinicopathological Characteristics of the Patients
From July 2003 to July 2013, a total of 189 patients underwent simultaneous CRC and hepatic resection; 35 patients were excluded because of benign hepatic disease, incomplete material or the presence of other malignant tumors. Detailed clinicopathological data of the 154 patients who underwent simultaneous resection of SCRLMs are shown in Table 1. The majority of patients were male (57.8%) and younger than 60.0 years (66.2%). Most patients presented with a primary colon tumor (69.5%). The average number of metastases was 1.9761.30 (1.0-8.0). The average size of the largest metastasis was 3.8662.45 cm (0.5 cm-15 cm).
As shown in Fig. 1, Kaplan-Meier survival analysis revealed that the OS (Fig. 1a) and DFS (Fig. 1b) of CRC patients with bilobar LMs were significantly poorer than those patients with unilobar LMs (P,0.001; P,0.001, respectively), indicating a crucial impact of LMs distribution on clinical outcome in patients.

Overall Survival Analysis
Follow-up information was obtained for the 154 patients through July 2013. The median OS calculated from the time of diagnosis of the disease was 49 months. The 5-year OS rate was 46%. The median follow-up period was 36 months. At the last follow-up, 49 (31.8%) patients had died; 78 (50.6%) patients experienced tumor recurrence; 52 (35.6%) had recurrence in the liver only, 12 (8.2%) had recurrence in the lung only and 14 (9.6%) had recurrence in other sites. Among the 52 (35.6%) patients with recurrence in the liver only, 13 (25%) underwent a repeat R0 resection, and 39 (75%) received adjuvant chemotherapy. The former had a good prognosis, as determined with the Kaplan-Meier method (P = 0.013) (Fig. 2).
In univariate analyses, the following factors were significantly associated with poor OS: CEA level (.5 ng/ml), positive lymph node status, vascular invasion, nerve invasion and BRAF mutation in the primary tumor, number of metastases ($4), bilobar distribution, and non-R0 resection of LMs. Multivariate analysis indicated that positive lymph node status, vascular invasion, BRAF mutation, bilobar LM distribution and non-R0 resection of LMs were significantly associated with poor prognosis. The details of the univariate and multivariate analyses are shown in Table 2. Based on the results of multivariate analyses, the PI for OS was calculated as follows: PI = (1.4136positive lymph nodes) + (1.0746vascular invasion) + (0.9286BRAF mutation) + (0.8466bilobar LM distribution) + (1.9456non-R0 LM resection) (b values was shown in Table 2). Kaplan-Meier estimates were calculated according to the PI value, and the cut-off point was the third quartile. A high PI (.5.978) was found to be highly predictive of the short-term outcome (P,0.001) (Fig. 3a). The 5-year OS rate for patients with low PI (#5.978) was 55%, whereas the 5-year survival rate for patients with high PI (.5.978) was 0%.

Disease-free Survival Analysis
The 5-year DFS of the 154 CRLM patients was 35%. In univariate analyses, tumor deposits, BRAF mutation, number of LMs ($4), bilobar LM distribution and extrahepatic metastases resection were significantly associated with shorter DFS. Multivariate analysis indicated that tumor deposits, BRAF mutation and bilobar LM distribution were significantly associated with shorter DFS. The details of the univariate and multivariate analyses are shown in Table 3.
The PI for DFS was calculated as follows: PI = (1.2576BRAF mutation) + (0.8446bilobar LM distribution) + (0.7306tumor deposits) (b values was shown in Table 3). Kaplan-Meier estimates were calculated according to the PI value; the cut-off point was the third quartile. A high PI (.2.945) was found to be highly predictive of the short-term outcome (P,0.001) (Fig. 3b). The 5year DFS rate for patients with a low PI (#2.945) was 41%, whereas the 3-year survival rate for patients with a high PI (. 2.945) was 5%.

Subgroup Analysis for Colon and Rectal Cancer
In consideration of the heterogeneity between colon and rectal cancer, we performed the categorized analysis according to high and low PI value for colon and rectal cancer patients respectively. Kaplan-Meier estimates were calculated according to high and low PI value. In colon cancer patients, a high PI (.5.978) was found to be highly predictive of the short-term OS (P,0.001) (Fig. 4a) and also high PI (.2.945) was found to be highly predictive of the short-term DFS (P,0.001) (Fig. 4b). In rectal cancer patients, a high PI (.5.978) was found to be highly predictive of the short-term OS (P = 0.001) (Fig. 5a), but high PI (.2.945) was not found to be highly predictive of the short-term DFS (P = 0.145) (Fig. 5b).

Discussion
In recent decades, the strategy of delayed hepatectomy has gained popularity and has been established as the standard surgical practice. Some authors hold the view that simultaneous resections may increase the rate of postoperative complications [17]. Given that the morbidity and mortality associated with hepatectomy have decreased substantially over the past 20 years, the classic paradigm of a staged operation for SCRLMs has been questioned. Specifically, given the improved feasibility and safety of major hepatic resections, some investigators have suggested that the simultaneous approach to SCRLMs may be preferable. Other large studies, including meta-analyses [3,4,8,9], have also shown that simultaneous resections are not associated with increased rates of hepatic or colon complications compared with delayed resection. Finally, as demonstrated by some large studies and meta-analyses examining the timing of hepatectomy for patients with SCRLMs, long-term OS and DFS are similar between the simultaneous and delayed resection groups [3,4,18]. In our study, the median overall survival was 49 months, the 5-year OS rate was 46%, and the 5-year DFS was 35%. These findings are in agreement with present studies [3,4,8,9,11,19].
Recently, Abbott AM et al [18] found significant predictors of short outcomes for CRLMs about complications, mortality, and prolonged length of stay including age .70 years, male gender, nonprivate health insurance, and Elixhauser score. However, the most important problem is identifying determinants of the longterm outcome. Few studies have examined factors that are prognostic of OS and DFS following synchronous resection, and the sample sizes were very small [10,[19][20][21][22]. Most studies have used a Clinical Risk Score (including nodal status of the primary tumor, a disease-free interval from discovery of the primary tumor to discovery of LM,12 months, .1 tumor, a preoperative CEA level .200 ng/ml and size of the largest tumor .5 cm) [20] and the Basingstoke Predictive Index (including .3 hepatic metastases, node-positive primary disease, a poorly differentiated primary tumor, presence of extra-hepatic disease, tumor diameter $5 cm, CEA level .60 ng/ml and positive resection margins) for riskcategorizing patients who may benefit from intensive surveillance as well as for the selection of adjuvant therapy and trials [21]. However, there were more MCRLMs in these studies, and the prognostic factor ''discovery of liver metastasis of ,12 months'' is unsuitable for predicting recurrence in SCRLM patients after simultaneous resection, as the OS and DFS of SCRLM patients are shorter than those of MCRLM patients. Furthermore, additional emphasis is now focused on tumor molecular stratification. Thus, the Clinical Risk Score must be updated. In our study, we found significant independent predictors of short OS, including positive lymph node status, vascular invasion, BRAF  mutation, bilobar LM distribution and non-R0 resection of LMs. We determined the PI using these predictors, and a high PI (. 5.978) was found to be highly predictive of short OS (P,0.001).
We also performed categorized analysis using tumor deposits, BRAF mutation and bilobar LM distribution for DFS, and we found that a high PI (.2.945) was highly predictive of short-term DFS (P,0.001). We also performed the categorized analysis according to high and low PI value for colon and rectal cancer patients respectively. In colon cancer patients, a high PI was found to be highly predictive of the short-term OS and DFS. In rectal cancer patients, a high PI was found to be highly predictive of the short-term OS, but high PI was not found to be highly predictive of the short-term DFS, because of the small sample of rectal cancer.
In contrast to previous series, the number of LMs [10,[20][21][22] and the size of LMs [20,21] were not independent predictors in our study. We examined different cut-offs for number and size in the univariate analyses, and we found that the hazard ratio of number of LMs (#3; $4) and size of LMs (,5 cm; $5 cm) was maximal, in agreement with the above studies [10,[20][21][22]. Therefore, we preliminarily concluded that if R0 resection is achieved, the number and size of LMs do not conclusively influence OS; however, additional evidence is needed. In many studies, a high preoperative CEA has been associated with poor survival, particularly at higher cut-off levels [11,[19][20][21]. However, our study and other studies [10,22] have failed to show that CEA is a significant prognostic factor, which could partially be due to the use of different cut-off values. In fact, we examined different cut-off values in univariate analyses and found that the hazard ratio of CEA .5 ng/ml/,5 ng/ml was maximal.
We first found that bilobar LM distribution and BRAF mutation of the primary tumor are independently associated with worse OS and DFS, respectively. Very few studies have explored the predictive role of bilobar distribution. In our study, OS and DFS of CRC patients with bilobar LMs were significantly poorer than those patients with unilobar LMs, indicating a crucial impact of LMs distribution on clinical outcome in patients.
KRAS and BRAF, which function downstream of the epidermal growth factor receptor (EGFR) signaling pathway, play important roles in the initiation and progression of CRC. Mutant KRAS is a predictor of resistance to EGFR monoclonal antibodies; thus, mutations in this gene are useful markers in clinical practice for patients with metastatic CRC [23]. Other studies have reported that BRAF mutations appear to be prognostic rather than predictive, as patients with metastatic CRC who do not receive cetuximab also have markedly reduced survival when their tumors harbor a BRAF mutation [24,25]. Therefore, it is essential to include these two genes in studies attempting to identify prognostic predictors for SCRLM. In the current study, we determined that BRAF is an independent predictor of both OS and DFS.
Despite the theoretical advantages of neo-adjuvant chemotherapy, including the eradication of microscopic neoplastic foci, practical evidence for its benefit in patients with resectable SCRLM is weak. According to the results of the EORTC 40983 randomized trial, perioperative chemotherapy improves the outcome of resectable CRLM patients [26]; however, the control group in this trial only underwent surgery without adjuvant chemotherapy. Our study and some retrospective series specifically focused on resectable synchronous metastases have failed to demonstrate any survival advantage in patients receiving neoadjuvant treatments compared to adjuvant chemotherapy [27,28]. Reddy SK et al [27] suggested that chemotherapy administered after but not before SCRLM resection is associated with improved DFS and OS. Another study including patients with SCRLMs and MCRLMs demonstrated that the effects of pre-and post-operative chemotherapy are indistinguishable [29] and that chemotherapy administered after but not before resection is associated with a good prognosis [30]. Prospective randomized trials are needed to determine the optimal timing of chemotherapy.

Conclusions
Despite the limitations of the patient data showing a fair amount of early censoring, based on our preliminary data, we can make a conclusion that simultaneous resection of SCRLM may produce  Table 3. Univariate and multivariate analyses of associations between clinicopathological characteristics and disease-free survival after curative resection of liver metastases.   good long-term outcomes. Patients with low PI have longer OS and DFS, while those with high PI have shorter OS and DFS. Thus, patients with high PI may be provided more aggressive treatment and intensive surveillance, This model needs further validation.