Abstract
Background/Aim: Routine use of adjuvant chemotherapy (AC) following hepatectomy for colorectal liver metastases (CRLM) is not universally practiced because of the lack of supporting evidence. Therefore, we investigated the efficacy of AC following curative CRLM resection. Patients and Methods: Among the 742 patients who underwent their first hepatectomy for CRLM at our institution, 335 were stratified into surgery alone (SA; n=162) and AC (n=173) groups. Poor prognostic factors for SA were identified using multivariate logistic regression analysis. Propensity score matching was used to compare the clinical outcomes between SA and AC groups according to the number of prognostic factors. Results: Multivariate analysis showed that preoperative carcinoembryonic antigen (CEA) levels (≥10 ng/ml; p=0.01), primary lymph node metastases (≥1; p=0.0001), and the number (n≥4; p=0.01) and maximum diameter (≥5 cm; p=0.00001) of CRLM tumours were independent poor prognostic factors for overall survival (OS) in the SA group. Patients with ≥3 risk factors were categorized as being high risk. After propensity score matching, the 5-year OS rate was significantly higher in the AC group (n=13) than that in the SA group (n=15; 47.9% vs. 7.3%; p=0.03) among high-risk patients. Conclusion: Adjuvant chemotherapy after curative CRLM resection may improve the prognosis of patients with three or more risk factors including preoperative CEA levels ≥10 g/ml, primary lymph node metastases ≥1, number (≥4) and maximum diameter (≥5 cm) of CRLM tumours.
Colorectal cancer (CRC) is among the leading causes of cancer-related deaths worldwide (1). Approximately 20% of patients with CRC present with synchronous distant metastases, and another 20% develop metachronous metastases (1). The liver is the most common metastatic site for CRC (2), and hepatectomy is the basis of its treatment (3). The 5-year overall survival (OS) rate following curative hepatectomy has been reported to range from 45%-61% (4). However, not all patients who undergo hepatectomy have a favourable long-term prognosis. Most treatment failures occur due to local hepatic recurrences and/or lung metastases, with recurrence occurring within the first 2 years post-hepatectomy (5). The postoperative recurrence rate has been found to be approximately 75% (4). Considering these facts, the combination of resection and chemotherapy may reduce the risk of recurrence.
A consensus has been established for postoperative adjuvant chemotherapy (AC) using oxaliplatin combination therapy or fluorinated pyrimidine monotherapy for colon cancer cases with lymph node metastasis (6). However, routine use of AC following hepatectomy for colorectal liver metastases (CRLMs) is not universally practiced because of the lack of supporting evidence (5). Several randomised control trials have been conducted to resolve this clinical dilemma (3, 5, 7, 8). In the JCOG0603 study, recurrence-free survival and OS were compared after microscopically negative margin-free removal via surgery (R0) of the primary and liver metastases between the surgery alone (SA) group and the group undergoing surgery with postoperative AC using modified folinic acid, fluorouracil, and oxaliplatin. The median disease-free survival (DFS) was longer in the surgery with AC group than that in the SA group. However, there was no difference in OS between the two groups (7). Conversely, chemotherapy is used to prolong OS for unresectable CRLMs (9). Therefore, there may be cases in which AC is effective even after curative CRLM hepatectomy. The purpose of our study was to evaluate the efficacy of AC after curative resection of CRLM.
Patients and Methods
Study design. Among the 742 patients who underwent their first hepatectomy for CRLM in our institution from January 1991–December 2020, 335 cases were considered in this study. Patients with extrahepatic metastases or gross residual tumour (R2) following resection (n=136) and those who underwent preoperative chemotherapy (n=202), microscopically positive-margin (R1) resection (n=65), and liver-first surgery prior to primary resection (n=4) were excluded. Cases with an overlap of these conditions were also omitted. Of the 335 cases, 173 patients underwent AC therapy following surgery, and the remaining 162 patients were retained as the SA group. The prognostic factors influencing the OS rate of the SA group were identified using multivariate logistic regression analysis. In addition, the short- and long-term clinical outcomes were compared through propensity score matching between the AC and SA groups according to the number of poor prognostic factors. The median range of the follow-up period was 45 months (0.3-254.4 months). This retrospective study was approved by the Yokohama City University Ethics Committee (ethics committee number; B220300014) and was compliant with the Declaration of Helsinki. The requirement for written informed consent was waived owing to the retrospective nature of the study.
Clinicopathological characteristics. The patient-related clinicopathological variables analysed were age (<65 vs. ≥65 years), sex (male vs. female), and initial carcinoembryonic antigen (CEA) levels (<10 vs. ≥10 ng/ml). The primary tumour-related variables analysed included the primary lesion site (right vs. left), primary histological type (well/moderate vs. others), lymph node metastases (0 vs. ≥1), depth of tumour invasion (T3, 4 vs. Tis-T2), and lymphatic (0 vs. ≥1) and venous invasion (0 vs. ≥1). The liver metastasis-related variables analysed included the number of metastatic tumours (1-3 vs. ≥4), maximum diameter (<50 vs. ≥50 mm), appearance time (synchronous vs. metachronous), and tumour distribution (unilobar vs. bilobar). The treatment-related variables analysed were staged hepatectomy (performed vs. not performed) and chemotherapy after primary resection (administered vs. not administered). In addition, propensity score matching was performed to minimise the differences in baseline characteristics between AC and SA. The propensity score for each patient was estimated by logistic regression analysis using the above-mentioned variables.
Adjuvant chemotherapy. Adjuvant chemotherapy (via hepatic arterial, portal vein infusion, or systemic administration) with fluorouracil, folinic acid, and oxaliplatin or irinotecan had been considered for all patients who underwent hepatectomy (10). However, since only a few studies have shown a survival benefit (3, 5), we considered whether to administer AC in individual cases from 2019 to date.
Hepatectomy. The primary method of hepatectomy used was non-anatomical hepatectomy with negative surgical margins, and anatomical hepatectomy was performed only if it was advantageous in terms of complete resection (R0), operative time, blood loss, or invasiveness. Portal vein embolization or two-stage hepatectomy was planned when the remnant prognostic score was low based on volumetry, the indocyanine green retention rate, and patient age (11). In addition to confirming the absence of residual tumours in the remnant liver, intraoperative ultrasonography was performed in all cases to detect occult tumours undetected by preoperative imaging and confirm the anatomical relationships between tumours and vasculo-biliary structures. R0 resection was considered complete when tumour-free resection margins were confirmed through pathological findings.
Outcomes. Overall survival was defined as the time from hepatectomy until death from any cause, disease-free survival was defined as the time from hepatectomy until first recurrence, and synchronous CRLM was defined as the presence of metastases in the liver at the time of primary CRC diagnosis.
Follow-up. Patients were assessed for recurrence following hepatectomy using contrast-enhanced computed tomography (CT; every 4-6 months), blood tests, and tumour markers (every 2-3 months). When recurrence in the remnant liver was suspected, magnetic resonance imaging was performed, and the appearance of new lesions was further investigated. Extrahepatic recurrence in the chest and pelvis was determined using CT. In addition, fluorodeoxyglucose positron emission tomography was sometimes performed to detect other distant metastases. Recurrence was adjudicated when imaging studies showed new lesions with typical features of CRC/CRLMs compared with those from previous imaging findings. Where applicable, recurrent CRLMs were treated by repeated resections, and when there was no indication for resection, chemotherapy, radiotherapy, radiation, or palliative care was chosen.
Statistical analysis. Quantitative variables were expressed as medians (interquartile ranges), and categorical variables as numbers and percentages. Continuous data were compared using the Mann–Whitney U-test and categorical data using the chi-square test. Survival curves were drawn using the Kaplan–Meier method and compared using the log-rank test. Statistically significant variables in the univariate analysis were included in the multivariate analysis. Stepwise logistic regression was used to perform multivariate analysis. All statistical analyses were performed using SPSS Base 11.0 J (SPSS Inc, Chicago, IL, USA). A p-value <0.05 was considered statistically significant.
Results
Baseline characteristics before propensity score matching. Before propensity score matching, there were 162 patients in the SA group and 173 patients in the AC group. Baseline characteristics that significantly differed between the two groups included age (≥65 years, p=0.01), primary tumour location (right, p=0.04), and lymphatic invasion (≥1, p=0.02) (Table I). The AC regimens used for treatment in this phase of the study were fluorinated pyrimidine- (n=143), oxaliplatin- (n=25), and irinotecan-based regimens (n=5). Adjuvant chemotherapy was started at a median of 46 days after surgery (46-146 days), and its median duration was 5.3 months (0.03-48 months).
Baseline characteristics before propensity score matching between the surgery alone (SA) and adjuvant chemotherapy (AC) groups.
Prognostic factors for SA group. In the univariate analysis, preoperative CEA levels (≥10 ng/ml, p=0.01), primary lymph node metastases (≥1, p=0.00001), and the number (≥4, p=0.01) and maximum diameter (≥5 cm, p=0.00001) of metastatic liver tumours were associated with poor OS in the SA group. In the multivariate analysis, preoperative CEA levels [hazard ratio (HR)=2.164, 95% confidence interval (CI)=1.098-4.267; p=0.02], primary lymph node metastases (HR=6.183, 95%CI=2.390-15.994; p=0.0001), and the number (HR=2.216, 95%CI=1.074-4.570; p=0.03) and maximum diameter (HR=2.478, 95%CI=1.285-4.777; p=0.007) of metastatic liver tumours were independent prognostic factors (Table II).
Prognostic factors for surgery alone (SA) patients.
Five-year OS rates of patients possessing zero (n=37), one (n=61), two (n=44), three (n=16), and four (n=4) risk factors were 90.1, 80.6, 58.9,13.2, and 0%, respectively. In our study, we classified patients with ≥3 and ≤2 risk factors into high- and low-risk categories, respectively. The OS and DFS rates of SA patients in the high-risk category (n=20) were significantly worse than those in the low-risk category (n=142; p=0.00001 and p=0.001, respectively; Figure 1A and B).
Kaplan–Meier curves of (A) overall survival and (B) disease-free survival stratified by risk. Patients in the high- (20 patients) and low- (142 patients) risk groups are represented by thin and thick lines, respectively.
Baseline characteristics after propensity score matching. In the SA and AC groups, 128 patients were selected by propensity matching. After matching preoperative baseline characteristics, perioperative factors were also comparable between the two groups (Table III). The AC regimens used for treatment in this phase of the study were fluorinated pyrimidine- (n=109), oxaliplatin- (n=15), and irinotecan-based regimens (n=4). The routes of administration were hepatic artery or portal vein infusion (n=42), systemic (n=54), and hepatic artery and systemic (n=32). In the SA group, 15 patients met the high-risk criteria, and 113 met the low-risk criteria, whereas in the AC group, 13 met the high-risk criteria, and 115 met the low-risk criteria (Table III). The clinical backgrounds shown in Table III were comparable when stratified by high- and low-risk categories.
Clinicopathological characteristics after propensity score matching.
Clinical outcomes after propensity score matching. In the short-term outcomes, the amount of intraoperative bleeding and frequency of red blood cell transfusions were significantly higher in the AC group than those in the SA group (p=0.0001 and p=0.04, respectively). However, there was no significant difference in the frequency of postoperative complication occurrence and length of postoperative hospital stay between the two groups (p=0.47 and p=0.06; Table IV).
Short-term results in the surgery alone (SA) and adjuvant chemotherapy (AC) groups after propensity score matching.
In the long-term outcomes, there was no significant difference in the OS (p=0.67; Figure 2A) and DFS rates (p=0.69; Figure 2B) between the two groups in the low-risk category. Conversely, in the high-risk category, the OS rate of the AC group was significantly higher than that of the SA group (p=0.03; Figure 3A). The median recurrence interval of the AC group tended to be prolonged compared with that of the SA group, though not significantly (15.2 vs. 5.0 months; p=0.05; Figure 3B).
Kaplan–Meier curves of (A) overall survival and (B) disease-free survival in low-risk patients in the surgery alone (SA) (thin line; 115 patients) and adjuvant chemotherapy (AC) (thick line; 113 patients) groups after propensity score matching.
Kaplan–Meier curves of (A) overall survival and (B) disease-free survival in high-risk patients in the surgery alone (SA) (thin line; 15 patients) and adjuvant chemotherapy (AC) (thick line; 13 patients) groups after propensity score matching.
Among the patients in the high-risk category, recurrence after hepatectomy was observed in 14 (93.3%) and 10 (76.9%) patients of the SA and CA groups, respectively, and the difference was not statistically significant (p=0.21). The remnant liver was found to be the most common site of initial recurrence in both groups, with no significant difference between them. However, simultaneous recurrence in two or more organs was significantly more frequent in the SA group than in the AC group (50% vs. 10%; p=0.03). Treatment for initial recurrence, including resection or chemotherapy, was significantly more frequently adopted in the AC group than in the SA group (100 vs. 57.1%; p=0.01) (Table V). The median survival time after recurrence in the high-risk category was slightly longer in the AC group than that in the SA group (22.2 and 19.9 months; p=0.56). There was one long-term survivor after recurrence in the AC group. In this case, tegafur/uracil and leucovorin were administered as AC for 6 months after hepatectomy because of its less toxicity and benefit to achieve tumour control (12). However, single-organ recurrence (lymph node) occurred 21.1 months after hepatectomy. Thereafter, chemotherapy (capecitabine and oxaliplatin+bevacizumab) was administered, and the patient survived for 5 years after recurrence.
Recurrence patterns and treatment after propensity score matching.
Discussion
This study revealed a significantly worse OS rate of patients in the high-risk category compared with that in low-risk category following curative CRLM resection. In our initial CRLM hepatectomy population, 162 (12.3%) high-risk cases were identified before matching and approximately 10% benefited from AC. The novelty of this study is that the efficacy of AC following curative CRLM resection was demonstrated after risk stratification and propensity score matching, and the outcomes after recurrence were investigated in detail.
Various prognostic factors for CRLMs, including node positivity, poor differentiation or tumour location of CRC, number of hepatic metastases >3, tumour diameter ≥5 cm, CEA >60 ng/ml, disease-free interval <12 months, short doubling time of liver metastases, and synchronous liver metastases have also been reported (13-18). Patients with these poor prognostic factors have a poor prognosis even with curative resection. In other words, AC is not effective for all patients, and the effectiveness of AC should be expected only for patients in the high-risk category.
Some studies on randomised controlled trials of AC after CRLM resection demonstrated an extension in the DFS period, although there was no OS extension (3, 5, 7). Several reasons have been identified for this finding (7). Primarily, AC may affect the liver parenchyma and potentially limits the process of neovascularization of the liver (19). As a result, the detection of liver recurrence may be delayed. The timing of detection may be affected due to oxaliplatin-induced sinus disorders (20), fluorouracil-induced liver parenchymal disorders (21), and irinotecan-induced fatty degeneration (22). Additionally, there may be an adjuvant therapy-related shortening of survival, with AC having eliminated the less malignant tumour cells, leaving only the most aggressive clones behind (23). Since these remaining clones were likely exposed to adjuvant systemic treatment previously, re-exposure would be less effective, resulting in treatment-induced resistance (23).
Nevertheless, our results contradicted these studies, showing a significantly superior OS in the AC group and slightly prolonged DFS, which could be attributed to the administration of a treatment regimen after recurrence (3, 19, 24). When treating cases of recurrence after hepatectomy, resection of the intra- and extrahepatic recurrent sites is crucial for prolonging patient survival (25), with a 5-year OS of 40%-60% (25, 26). Even when chemotherapy is administered to treat recurrence, adding irinotecan or oxaliplatin to the treatment regimen could result in a 14-20.3-month increase in the median survival (27-29), which differs from the 7.5-month median survival of untreated patients (30). The present study demonstrated that, although more than half of the patients in both the high-risk AC and SA groups relapsed after the first hepatectomy, resection or chemotherapy that was expected to have therapeutic effects as a treatment for recurrence was selected more frequently in the AC patients than in the SA patients.
As proof of this finding, the number of organs initially affected by recurrence was significantly lower in the AC group than that in the SA group, thereby influencing the difference in treatment choices following recurrence. Supportive care or radiation was frequently selected in the SA group, which encountered multiple organ recurrences. AC may be used to control a small metastatic lesion, resulting in fewer organs being affected by recurrence. This phenomenon probably contributed to OS prolongation in the high-risk AC group. In contrast, another retrospective case-matched study demonstrated that both OS and DFS improved in patients with synchronous CRLMs in the AC group (31). However, since the treatment of recurrence cases was not investigated in-depth, the exact effect of AC could not be established.
This study has several limitations. First and foremost, the present study was conducted at only two centres with a small sample size. Second, its retrospective nature introduces the inevitable risk of selection bias, which could not be completely eradicated, despite using propensity score matching to reduce confounding by indication. Third, molecular biological factors, such as reticular activating system status and microsatellite instability, are considered prognostic (32, 33). However, information on these factors was unavailable during the present study. Despite these drawbacks, our findings from the SA group indicate that AC may be effective for treating high-risk category patients.
In conclusion, multidisciplinary treatment, including AC after curative CRLM resection, may improve the prognosis of patients that have three or more risk factors including a preoperative CEA (≥10 g/ml), primary lymph node metastases ≥1, number of metastatic liver tumours ≥4, and maximum diameter of metastatic liver tumours ≥5 cm. We recognise the urgent need for future prospective and/or multi-institutional studies with larger samples to verify our findings.
Acknowledgements
The Authors wish to thank Editage (www.editage.com) for their writing support.
Footnotes
Authors’ Contributions
Takeda K, Kikuchi Y, Sawada Y, Kumamoto T, and Watanabe J performed the research; Takeda K drafted the manuscript; Kunisaki C, Misumi T, and Endo I revised the manuscript critically; all Authors read and approved the final manuscript.
Conflicts of Interest
The Authors have no conflicts of interest to declare in relation to this study.
- Received September 2, 2022.
- Revision received September 13, 2022.
- Accepted September 26, 2022.
- Copyright © 2022 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
