Introduction

Surgical resection of colorectal liver metastasis (CLM) is accepted as the only potentially curative treatment. However, despite advances in the management of CLM, recurrence is found in the liver remnant or at other sites in 50–75% of patients who undergo liver resection for CLM [1,2,3,4]. It is thought that the high frequency of recurrence after liver resection is due to occult metastases derived from the primary colorectal cancer (CRC) and residual lesions scattered from CLM [5, 6].

The serum levels of carcinoembryonic antigen (CEA), a glycoprotein, frequently increase in patients with various types of cancer. Thus, CEA testing is widely used in the management of CRC patients, especially as a surveillance tool after the resection of primary stages I–III CRC tumors, as an early detector of recurrence [7,8,9,10]. Moreover, many studies have shown that perioperative CEA is a predictive marker of recurrence after complete (R0) resection for metastatic CRC [11,12,13,14]. However, it is unclear how perioperative CEA should be used in the management of patients with CLM and the efficacy of CEA surveillance after surgery for CLM has not been fully investigated. If the serum CEA level after liver resection reflects the existence of residual occult tumors, an elevated CEA level could be not only a predictor of recurrence, but also a good surveillance tool after liver resection. We conducted this study to define the association of perioperative CEA levels and the characteristics of recurrence after liver resection and to identify the possible roles of CEA testing in the management of patients after R0 resection for primary tumors and then CLM.

Patients and methods

Study design

We reviewed medical records collected from the 20 institutes participating in the Japanese Study Group for Postoperative Follow-up of Colorectal Cancer (JFUP-CRC). The JFUP-CRC is a joint study group, established in 2001 to propose the most adaptive and effective follow-up program for the management of CRC patients (the investigators in this group are listed in Acknowledgements).

To assess the value of CEA as a predictor of recurrence or as a surveillance tool after liver resection, we reviewed the collected data and analyzed the association between perioperative CEA levels and the characteristics of recurrence after liver resection for synchronous CLM. All patients enrolled in this study underwent surgical procedures that were covered by the Japanese national health insurance. This study was exempt from institutional ethical review because patient information could not be identified from the data.

Data collection

The subjects of this retrospective analysis were patients who underwent R0 resection for both primary tumors and synchronous CLM between January, 1997, and December, 2007. Patients with extrahepatic metastases were excluded. We collected data on the characteristics of the primary tumor and liver metastases, perioperative chemotherapy, recurrence, and survival, as well as the preoperative and postoperative CEA levels, defined as levels measured within 1 month before and within 3 months after liver resection, respectively. The CEA level at the time of recurrence was also measured. Patients whose CEA levels had not been measured were excluded from the study population. Patients who received any chemotherapy for more than 6 months after liver resection were regarded as having received adjuvant chemotherapy. All other patients were grouped as having received ‘no adjuvant chemotherapy.’

Statistical analysis

Categorical variables were compared using the Chi-square test. Multivariate analysis was completed for factors with a P value <0.10 on univariate analysis using a logistic regression model. To evaluate the survival time and the time until recurrence, the Kaplan–Meier method was used and a comparison was made using the log-rank test. Differences with a P value of <0.05 were considered significant. All statistical analyses were performed using SPSS version 13.0 (SPSS Inc., Chicago, IL, USA).

Results

Patient characteristics and clinical outcomes

We reviewed the records of 604 patients with synchronous CLM who underwent resection of primary tumors and CLM between 1997 and 2007. The median follow-up after liver resection was 52 months (range 3–199 months). Table 1 summarizes the clinical and pathological data of the 604 patients. There were 383 (63.4%) men and 221 (36.6%) women, with a mean age of 63 years (range 29–91 years). The primary tumors were advanced with regard to invasion [T3 and T4, n = 408 (67.5%)] and nodal status [node positive, n = 431 (71.4%)], and 40.7% of tumors were located in the rectum. The median diameter of the largest metastasis was 25 mm (range 4–170 mm), and the median number of metastases was 2 (range 1–32). Among the 604 patients, 470 (77.8%) underwent liver resection simultaneously to the primary tumor resection, and the remaining patients underwent metachronous liver resection after resection of the primary tumor. Fifty (8.3%) patients received neoadjuvant chemotherapy, and 364 (60.3%) received adjuvant chemotherapy. Recurrence was detected in 445 (73.7%) patients during the median follow-up time of 52 months and within 3 years after liver resection in 91.0%. The median survival time after liver resection was 66 months (range 3–199 months), and the 5-year cumulative overall survival rate was 51.9%.

Table 1 Patient characteristics and analysis of factors associated with recurrence after liver resection
Full size table

CEA as a predictor of recurrence

To assess the predictive factor of recurrence, clinicopathological variables were compared between patients with and without recurrence (Fig. 1). We investigated which of the following four CEA variables was the strongest predictor of recurrence: a high preoperative CEA level; a high postoperative CEA level; a high relative CEA level, being the ratio of postoperative CEA against preoperative CEA; and both high preoperative and postoperative CEA levels. The cutoff levels for preoperative, postoperative, and relative CEA were calculated using receiver operating characteristic (ROC) curves (Supplementary Fig. 1). The most appropriate cutoff levels were 90 ng/ml for preoperative CEA, 5 ng/ml for postoperative CEA, and 0.75 for relative CEA, respectively. Those four variables were entered individually into the univariate analysis (a preoperative CEA level > or ≤90; a postoperative CEA level > or ≤5; a relative CEA level > or ≤0.75; both a preoperative CEA level >5 and a postoperative CEA level >5 or not), and each one was entered separately into the multivariate analysis with the other predictors (Table 1). A postoperative CEA level >5 ng/ml was identified as an independent predictor of recurrence with the highest HR (2.25, 95% confidence interval (CI) 1.29–3.91, P = 0.004). The recurrence rates were 84.2% for patients with a postoperative CEA level >5 ng/ml and 70.5% for those with a postoperative CEA level ≤5 ng/ml (P = 0.001). For the prediction of recurrence, a postoperative CEA level >5 ng/ml had the highest specificity (86.2%) and the highest positive predictive value (PPV 84.2%), while the sensitivity was 26.3%.

Fig. 1
figure 1

Analysis flowchart

Full size image

Characteristics of recurrence according to postoperative CEA

The 445 patients with recurrence were divided into two groups according to their postoperative CEA level: a high-CEA group with a postoperative CEA level >5 ng/ml (n = 117) and a low-CEA group with a postoperative CEA level ≤5 ng/ml (n = 328; Fig. 1). We compared the time to recurrence after liver resection, the sites of recurrence, and the CEA level at the time of recurrence between the groups.

Time to recurrence after liver resection

The median time until recurrence was significantly shorter in the high-CEA group than in the low-CEA group (5.0 vs. 9.1 months, P < 0.0001). Recurrence was found within 6 months of resection in 55.9% of patients in the high-CEA group vs. 27% of patients in the low-CEA group (Fig. 2).

Fig. 2
figure 2

Cumulative recurrence rates after liver resection

Full size image

Sites and distribution of recurrence

Table 2 shows the sites and distribution of recurrence in each group. More patients in the high-CEA group had metastases at multiple sites than those in the low-CEA group (36.8 vs. 21.6%, P = 0.001). The liver remnant was the most common site of recurrence in both groups (68.4 and 68.0%). Lung and lymph node metastases were more frequent in the high-CEA group than in the low-CEA group (40.1 vs. 28.4%; P = 0.018, and 17.1 vs. 10.4%; P = 0.056, respectively). There were no differences in recurrence rates at other sites.

Table 2 Sites and distribution of recurrence
Full size table

CEA level at the time of recurrence

A CEA level of >5 ng/ml was found at the time of recurrence in 52.7% of the patients in the low-CEA group. This means that the sensitivity of CEA testing to detect actual recurrence is 52.7%. We also measured the ratio of CEA at the time of recurrence vs. the postoperative CEA level in the high-CEA group. The median ratio was 1.25, with 41% of patients having a ratio of more than 1.5 and 52.1% having a ratio of more than 1.2 (Table 3). The specificity and accuracy of CEA testing to diagnosis actual recurrence are not available due to the lack of corresponding CEA levels in the patients without recurrence.

Table 3 Change in carcinoembryonic antigen levels at the time of recurrence
Full size table

Discussion

In accordance with the findings of previous studies that postoperative CEA is a predictive marker of recurrence [12, 13], the patients in this series with a postoperative CEA level >5 ng/ml were at a significantly higher risk of recurrence, with a shorter time until recurrence and a higher frequency of multiple metastatic sites. Our findings support the consensus that the serum CEA level after liver resection reflects the existence of residual occult tumors in the liver remnant or other organs.

An intensive adjuvant chemotherapy regimen after liver resection would be recommended for patients with a high recurrence risk. Although many studies [15,16,17,18] have investigated the efficacy of adjuvant chemotherapy after curative resection of CLM, effective regimens and indications for adjuvant chemotherapy have not yet been established. Since a postoperative CEA level >5 ng/ml had a high specificity and PPV to predict recurrence, this cutoff level could be used to stratify candidates who would benefit from adjuvant chemotherapy. From the biological viewpoint that a high postoperative CEA level may indicate residual occult tumors, this variable would be reasonable to use in selecting those candidates. However, the sensitivity of a postoperative CEA level >5 ng/ml was low, and even patients with a low postoperative CEA level (≤5 ng/ml) had a high recurrence rate of 70.5%. Therefore, additional criteria for selecting candidates for adjuvant chemotherapy will be necessary. Nodal metastases and a larger number of liver metastases, which were also independent predictors of recurrence with a high hazard ratio in this study, could be helpful.

An adequate and effective follow-up program to detect recurrent tumors earlier is important. Several meta-analyses [19,20,21] have provided evidence that intensive follow-up programs after curative surgery improve the outcome of patients with non-metastatic CRC. In particular, CEA testing, as well as imaging modalities, contributes to improving outcomes [22]. However, the efficacy of CEA testing as a surveillance tool after resection of metastatic CRC has not been fully investigated. Some guidelines recommend follow-up programs comprised of CEA testing and CT for patients who have undergone R0 resection for metastatic CRC. The ESMO Clinical Practice Guidelines [23] recommend a follow-up program with CEA testing and CT at intervals of 3–6 months during the first 3 years. Similarly, the NCCN Guidelines (Version 2. 2015) recommend a follow-up program of CEA testing and CT scans every 3–6 months for the first 2 years, then CEA testing every 6 months and CT scans every 6–12 months for the following 3 years. In the present study, almost half of the patients with a low postoperative CEA level did not have elevated CEA levels at the time of recurrence, suggesting that CEA monitoring is not useful in follow-up. Accordingly, intensive imaging studies as well as CEA monitoring would be necessary for earlier and more accurate detection of recurrent tumors.

This study was limited by its retrospective design; therefore, further prospective studies are warranted to confirm our findings. More data on chronological change in CEA levels after liver resection are also needed, and other factors that influence CEA elevation, such as smoking, should be taken into consideration, to clarify the role of CEA testing.

In conclusion, a postoperative CEA level >5 ng/ml was an independent and strong predictor of recurrence, which could assist in decisions about treatment after liver resection. However, since CEA monitoring was not reliable as a surveillance tool, intensive imaging studies based on the characteristics of recurrence would be necessary for optimal follow-up after liver resection for CLM.