Introduction

Liver is the most common metastasis site of colorectal cancer [1], with the metastasis rate of 40–50% [2]. Colorectal liver oligometastases (i.e., limited metastases confined to the liver) have been identified to benefit the survival of patients through resection and focal therapies [3, 4]. The 5-year survival rate of patients with oligostastatic colorectal liver metastases (CRLM) can be improved from 9% to 20%–50% by curative surgical resection [5,6,7,8].

Parenchymal-sparing hepatectomy (PSH) is recommended in patients with CRLM because microscopically positive surgical margins and surgical procedures have no effect on overall survival, consequently offering a high rate of repeat resection for liver recurrence [9,10,11]. In previous studies, PSH included non-anatomic resection (NAR) (including non-anatomic wedge resections) and anatomic resection (AR) (including anatomic segmentectomies and anatomic bisegmentectomies) [10,11,12,13]. Although NAR maintains functional liver remnants and reduces the risk of liver failure, it has the potential to increase the risk of intrahepatic recurrence compared to AR [14, 15]. Unfortunately, approximately half of patients have liver recurrence within 3 years after NAR [16]. Hence, following the principle of PSH, the choice of the appropriate surgical procedure for each lesion to balance between reducing intrahepatic recurrence and maintaining hepatic functional reserve demands further exploration.

Peritumoral pathological and imaging features including vascular invasion, bile duct invasion, peritumoral early enhancement, peritumoral hypointensity on hepatobiliary phase (HBP), and biliary duct dilatation probably predict a worse prognosis [17,18,19]. However, it is unclear whether these radiological risk factors can predict the long-term prognosis of different surgical procedures.

Thus, the aims of this study were to determine the impact of AR and NAR on the outcome for CRLM according to the principle of PSH and to investigate whether peritumoral imaging features to the CRLM could be used to predict long-term prognosis in different surgical procedures.

Materials and methods

Patients

This retrospective study was approved by our institutional review board. The database of our institution was reviewed from August 2014 to May 2022 inclusively to identify all patients with CRLM who had undergone preoperative (within 2 weeks before surgery) abdominal gadoxetic acid-enhanced magnetic resonance imaging (MRI) and curative surgery for CRLM, which was defined as hepatectomy with macroscopic clear resection margin[20]. Exclusion criteria: (a) patients with neoadjuvant chemotherapy or hepatic recurrence during adjuvant chemotherapy after primary site resection; (b) patients that received interventional therapy of liver before surgery, including transarterial chemoembolization or radiofrequency ablation; (c) patients that had coexistence of extrahepatic metastases; (d) patients with redo liver resection;(e) patients combined with malignancies at the other sites; (f) patients with non-PSH (≥3 segments) or combined resection (simultaneous AR and NAR)[12]; (g) patients with less than 12 months of follow-up; (h) patients with R1(microscopically positive surgical margins) surgical margin status; (i) patients with five or more CRLMs. The patient flowchart is shown in Figure 1.

Fig.1
figure 1

A flow diagram of the study population. CRLM colorectal liver metastasis, PSH parenchymal-sparing hepatectomy, AR anatomic resection, NAR non-anatomic resection.

Full size image

Clinicopathological characteristics were collected, including age, sex, virus status, primary tumor location, synchronous or metachronous liver metastasis, preoperative liver function tests (including galanine transaminase, aspartate transaminase, gamma glutamyltransferase, lactate dehydrogenase, and phosphatase alkaline) [21], serum carcinoembryonic antigen (CEA) level, serum alpha-fetoprotein (AFP) level, stage of the primary tumor, and type of hepatectomy. The liver function tests were classified as normal or abnormally elevated, according to the laboratory ranges [21].

Preoperative assessment and surgical procedure

Hepatectomy was indicated for cases in which all tumors could be removed with clear margins, leaving future liver remnant >30% of the total liver volume. Tri-phase contrast-enhanced computed tomography (CT) and abdominal gadoxetic acid-enhanced MRI were performed for tumor-staging. The preoperative estimate whole liver volume and the postoperative estimate remnant liver volume were calculated using tri-phase contrast-enhanced CT scans. Liver function was evaluated in terms of the indocyanine green retention rate at 15 minutes and CT volumetry. All surgical procedures were designed by the multidisciplinary team (MDT) discussion depending on clinical history, physical examination, serum laboratory tests, the number and location of the tumors, the probability of achieving a negative surgical margin, the need to preserve an adequate liver remnant, and relation of the tumor to vascular structures.

All surgical procedures were performed by two hepatobiliary surgeons with more than 20 years’ experience in hepatectomy who are members of the Colorectal Cancer MDT for hepatic resection of CRLM. Following laparotomy or laparoscopy, intraoperative ultrasonography examination of the liver was routinely performed to confirm the exact location and number of CRLM, while the relationship of the CRLM to the portal vein and hepatic vein was considered to guide resection. The type of resection (AR vs. NAR) was determined by the operating surgeon using a combination of preoperative and intraoperative evaluation.

MR imaging examination

All MR images were performed with a 3-T scanner (Signa HD Excite; GE Healthcare, Milwaukee, WI). Routine in-phase and opposed phase T1-weighted images (time of repetition [TR], 260 ms; time of echo [TE], 2.3 ms and 4.6 ms, respectively; flip angle 80°; matrix, 384×160; field of view, 400×400 mm; section thickness, 5 mm; intersection gap, 2 mm) were obtained. Pre-contrast images were obtained in a transverse plane with a fat-suppressed three-dimensional (3D) T1-weighted liver acquisition with volume acceleration-extended volume (LAVA-XV) sequence (TR, 4 ms; TE, 1.9 ms; flip angle 12°; matrix, 320×224; field of view, 380×304  mm; section thickness, 4 mm; intersection gap, 0 mm). All patients were given 25 mmol/kg (0.1 mL/kg) of gadoxetic acid (Gadoxetic Acid Disodium Injection; Chiatai Tianqing Pharma, Jiang Su, China) as an intravenous bolus, using a power injector at a rate of 1 mL/s, followed by a 20-mL saline flush. After the contrast administration, the early arterial phase (20–25s), the late arterial phase (40–45s), the portal venous phase (65–70s), transition phase (3–5min), and HBP (10–20min) images were obtained using a T1-weighted 3D LAVA-XV sequence. T2-weighted fast spin echo sequences (TR, 4400 ms; TE, 85 ms; flip angle 90°; matrix, 320×224; field of view, 400×300 mm; section thickness, 5 mm; intersection gap, 1 mm) were obtained using a respiratory-triggered technique.

Image analysis

All images were retrospectively and independently reviewed by two abdominal imaging radiologists (with 10 and 15 years of experience, respectively), who were blinded to the clinical and pathologic findings. In case of disagreements, adopt the decision through consultation. The following tumoral features were evaluated: number, size (maximum diameter on HBP), tumor shape (regular or irregular such as lobulated), peritumoral early enhancement, peritumoral hypointensity on HBP, and bile duct dilatation. Peritumoral early enhancement was evaluated on early and/or late arterial phase images and excluded peritumoral rim enhancement (Fig. 2a). Bile duct dilatation was defined as a peritumoral linear or branched hyperintensity area on fat-suppressed T2-weighted images (Fig. 2b). Peritumoral hypointensity on HBP was defined as wedge-shaped hypointense area of hepatic parenchyma located outside of the tumor margin on HBP (Fig. 2c). In patients with multiple CRLMs, the patient was included in the positive group if at least one tumor showed peritumoral early enhancement, peritumoral hypointensity on HBP, and bile duct dilatation.

Fig. 2
figure 2

Images in patients with colorectal liver metastases. (a) Contrast-enhanced late arterial phase T1-weighted magnetic resonance (MR) image shows a hypointense mass with peritumoral wedge-shaped enhancement (arrow). (b) Axial fat-suppressed T2-weighted MR image shows a strong linear hyperintensity (arrow). This indicates bile duct dilatation. (c) Contrast-enhanced 20-minute hepatobiliary phase MR image shows a hypointense mass with peritumoral wedge-shaped intermediate hypointensity (arrow).

Full size image

We also evaluated CRLM location (deep or surface) and distance from CRLM to vascular (<1mm or ≥1mm), given the possible impact on the choice of surgical procedure and prognosis. We identified patients who had deep-placed CRLMs whose margin was located >30 mm from the liver surface and others with surface-placed. Distance from CRLM to vascular was defined as the shortest distance from the tumor margin to the first-and second-order branches of the portal veins, hepatic veins, or hepatic arteries. In patients with multiple CRLMs, we recorded the deepest tumors and those nearest to the vascular.

Histologic analyses

All histologic specimens were analyzed by a pathologist (10 years of experience in gastrointestinal pathology), who was blinded to the original pathology reports, clinical data, imaging findings, and follow-up data. The surgical margin status and the presence or absence of portal vein, hepatic vein and bile duct invasion were re-evaluated using light microscopy for each patient. Surgical margin status was classified as R0 (microscopically negative surgical margins) or R1 (microscopically positive surgical margins) [22]. Portal vein, hepatic vein, and bile duct invasion were considered present when tumor cells were seen within the portal vein, hepatic vein, or bile duct channels in hematoxylin-eosin-stained sections [22].

Follow-up and adjuvant chemotherapy

Postoperative follow-up of patients was performed every 3-6 months during the first 2 years and every 6–12 months thereafter. The routine follow-up included contrast-enhanced CT or MR examinations and tumor marker (CEA and carbohydrate antigen 19-9) testing. Tumor recurrence was identified according to the imaging findings (CT or MRI). Liver recurrence-free survival (LRFS) is defined as the time from liver resection to intrahepatic recurrence, irrespective of the presence of additional recurrences in other organs. Recurrence-free survival (RFS) is defined as the time from liver resection to any disease recurrence (i.e., intrahepatic recurrence and extrahepatic metastases). Overall survival (OS) is defined as the interval between the operation and the date of any cause of death. All the cases without end events for each prognostic outcome were censored at the date of the last follow-up.

Postoperative chemotherapy was administrated following the standard National Comprehensive Cancer guidelines. The postoperative chemotherapy was based on FOLFOX (oxaliplatin, leucovorin, and 5-fluorouracil) and CapeOX (oxaliplatin and capecitabine).

Statistical analysis

Statistical analysis was performed with software (IBM SPSS, version 26.0. Armonk, NY: IBM Corp). If the data followed a normal distribution, mean ± standard deviation was used, whereas medians with interquartile ranges (IQRs) were reported if not. The chi-square test or the Fisher exact test were used to assess categorical variables. The unpaired 2-tailed t test or the Mann-Whitney U test were used to assess continuous variables, depending on the pattern of distribution. Interobserver agreement for peritumoral imaging features was assessed with kappa statistics. Interobserver agreement was defined as poor (κ < 0.20), fair (κ = 0.20–0.39), moderate (κ = 0.40–0.59), substantial (κ = 0.60–0.79), or almost perfect (κ = 0.80–1.00). Survival rates were estimated using the Kaplan-Meier method and were compared by using the log-rank test. Baseline variables that were considered clinically relevant or that showed a univariate relationship with outcome (p values less than or equal to 0.1 in the univariable analysis) were entered into multivariate Cox proportional-hazards regression model and used the automated backward elimination regression. Two-sided p < 0.05 was considered statistically significant.

Result

Patient characteristics

A total of 272 patients underwent hepatectomy for CRLM during the study period. One hundred and seventy patients were excluded, leaving 56 for analysis, including 16 who underwent AR and 40 NAR. The location and number of CRLMs on preoperative MR examination were consistent with those on intraoperative ultrasonography, with a total of 86 lesions, and all of them were resected. All 56 patients received postoperative chemotherapy, and the median number of completed chemotherapy cycles was 5 (range, 1–10). Baseline characteristics of the study population are summarized in Table 1. There were no statistical significances between the two groups in term of clinical-pathologic characteristics (all p > 0.05; Table 1).

Table 1 Baseline characteristics of patients
Full size table

Interobserver agreement between the two observers regarding peritumoral imaging features was almost perfect for bile duct dilatation (κ= 0.876; 95% confidence interval [CI] 0.707, 1.000), substantial for peritumoral early enhancement (κ = 0.709; 95% CI 0.491, 0.928) and peritumoral hypointensity on HBP (κ = 0.752; 95% CI 0.520, 0.984).

The median follow-up periods were not different between the AR (40 months, IQR: 22–81 months) and NAR groups (58 months, IQR: 41–62 months; p = 0.483). 50.0% of patients (28/56) and 66.1% of patients (37/56) developed liver recurrence and systemic recurrence, respectively. 37.5% of deaths (21/56) occurred during the entire follow-up period. Intrahepatic recurrence was significantly less common in the AR group [18.8% of patients (3/16) in the AR group and 62.5% of patients (25/40) in the NAR group (p = 0.003; Table 1)]. A similar systemic recurrence rate was observed in 62.5% of patients (10/16) in the AR group and 67.5% of patients in the NAR group (p = 0.721; Table 1). A similar mortality rate was observed in 31.3% of patients (5/16) in the AR group and 40.0% of patients (16/40) in the NAR group (p = 0.541; Table 1).

Correlation between imaging and pathologic features

The results showed that the correlation between imaging features (i.e., peritumoral early enhancement, peritumoral hypointensity on HBP, and bile duct dilatation) and pathologic invasion was not statistically significant (all p > 0.05, Table 2).

Table 2 Correlation of imaging features in peripheral region of tumor for pathologic features
Full size table

Predictors of LRFS, RFS and OS in the overall cohort

The 3-year LRFS rates were 78.6% and 36.6% in AR and NAR groups (p = 0.012), respectively. The 3-year RFS and OS rates were 41.7% and 81.3% in AR group, and 34.1% and 61.7% in NAR group (p = 0.794 and p = 0.302), respectively. Long-term prognosis of the AR group and NAR group is demonstrated in Figure 3.

Fig. 3
figure 3

Kaplan-Meier curves of liver recurrence-free survival (a), recurrence-free survival (b), and overall survival (c) of 16 patients who underwent anatomy resection and 40 patients who underwent non-anatomy resection.

Full size image

On multivariable analysis, NAR (p = 0.022; hazard ratio [HR] = 4.402; 95% CI 1.240,15.633), abnormal liver function (p = 0.014; HR = 4.071; 95% CI 1.327, 12.484) poorly differentiated primary tumor (p = 0.007; HR = 4.071; 95% CI 1.327,12.484) and two or more CRLMs (p = 0.035; HR = 2.675; 95% CI 1.073,6.667) were independently associated with worse LRFS, poorly differentiated primary tumor (p = 0.003; HR = 3.332; 95% CI 1.504,7.380) , primary tumor located in colon (p = 0.006; HR = 3.828; 95% CI 1.468,9.979) and two or more CRLMs (p = 0.042; HR = 2.183; 95% CI 1.029,4.629) were independently associated with worse RFS (Table 3). There was no independent predictor for OS (Table 3).

Table 3 Results of Cox proportional hazards model for identification of survival predictors in the overall cohort
Full size table

Predictors of LRFS, RFS and OS in the AR and NAR groups

In NAR group, on multivariate analysis, two or more CRLMs (p= 0.005; HR =3.862; 95% CI 1.512,9.864) and peritumoral early enhancement (p < 0.001; HR = 6.260; 95% CI 2.322,16.876) were independently associated with worse LRFS, poorly differentiated primary tumor (p = 0.029; HR = 3.505; 95% CI 1.139,10.781), primary tumor located in colon (p = 0.040; HR = 3.386; 95% CI 1.060,10.817), the largest tumor size of 5 cm or larger (p = 0.036; HR = 8.822; 95% CI 1.152,67.570) and peritumoral early enhancement (p = 0.035, HR = 2.516; 95% CI, 1.069,5.919) were independently associated with worse RFS, poorly differentiated primary tumor (p = 0.048; HR = 3.594; 95% CI 1.014,12.739) and the largest tumor size of 5 cm or larger (p = 0.001; HR = 70.315; 95% CI 5.567,888.188) were independently associated with worse OS (Table 4). In contrast, there were no independent predictors of LRFS, RFS and OS in AR group (Table 5). In AR and NAR groups, bile duct dilatation and peritumoral hypointensity on HBP were not independent predictors of LRFS, RFS and OS.

Table 4. Results of Cox proportional hazards model for identification of survival predictors in the NAR group
Full size table
Table 5 Results of Cox proportional hazards model for identification of survival predictors in the AR group
Full size table

Only 14 of 56 patients (5 of 16 and 9 of 40 in the AR and NAR group, respectively) presented with peritumoral early enhancement. All of 5 in the AR group had no intrahepatic recurrence and only 2 had systemic recurrence. However, all of 9 in the NAR group had intrahepatic recurrence. In NAR group, there was a significant difference in 3-year LRFS and RFS rates between patients with and without peritumoral early enhancement (LRFS: 11.1% vs. 44.4%, p = 0.001; RFS: 11.1% vs. 41.0%, p = 0.010; Fig. 4).

Fig. 4
figure 4

Liver recurrence-free survival (a) and recurrence-free survival (b) of 31 patients without peritumoral early enhancement and 9 patients with peritumoral early enhancement in non-anatomic resection (NAR) group.

Full size image

Discussion

Surgical resection can improve the survival of patients with CRLM, but the postoperative recurrence rate is high [8, 23]. On the basis of PSH principle, it remains a challenge to select the appropriate surgical procedure to reduce intrahepatic recurrence and maximize the functional liver remnant [12]. This study revealed that AR was associated with improved LRFS, although it did not correlate with either RFS or OS. Further, our analyses indicated that peritumoral early enhancement was a risk factor for LRFS and RFS in patients with NAR but not in patients with AR.

The influence of the surgical procedures on LRFS is undefined [11, 14, 15]. The underlying causes are the lack of clarity in the definition of AR, the failure to exclude patients receiving mixed AR and NAR, and the interchangeable use of the terms AR and non-PSH in the literature, which lead to confounding the results and limiting inter-study comparability [24]. We indicated that NAR was associated with an increased risk of liver recurrence compared with AR in patients undergoing PSH. Similar results have been reported by Finch et al. They also proposed that intrahepatic recurrence was significantly more common in the NAR group [25]. This could be because AR is associated with more extensive parenchymal resection, so coexisting micrometastases in the same lobe are removed [25]. Meanwhile, similar to previous studies, we also found that NAR did not present a disadvantage to the patients in terms of RFS and OS and poorly differentiated primary tumor was associated with worse RFS [25,26,27,28].

It was suggested that peritumoral imaging features with CRLM were predictors of long-term prognosis [19]. In our study, peritumoral early enhancement was associated with poor LRFS and poor RFS in the NAR group, although this predictive imaging feature was not demonstrated to be associated with a poor prognosis in the overall cohort. This is probably due to the fact that potentially ‘tumor-bearing’ portal tributaries cannot be removed by NAR [24, 29]. Thus, peritumoral early enhancement may be an important sign in surgical planning and may require resection of the potential "tumor-bearing" portal tributaries. Peritumoral early enhancement may be due to tumor compression of the surrounding hepatic parenchyma, portal veins and hepatic veins or portal vein obstruction caused by tumor cells, resulting in arterioportal shunt [30, 31]. Although there was a larger percentage of portal vein invasion in CRLM with peritumoral early enhancement, the difference did not reach statistical significance (p = 0.060). The study by Nakai et al. also did not demonstrate a correlation between peritumoral early enhancement and microvascular invasion, which was proved in hepatocellular carcinoma [19, 31, 32]. The reason may be that the imaging features and pathological features may not completely correspond, because our research is retrospective. Therefore, prospective research needs to be conducted to further explore the correlation between imaging features and pathological features.

Currently, it has been documented that biliary dilatation caused by liver metastases is usually the result of ductal invasion rather than biliary compression [33,34,35]. However, we did not find a correlation between bile duct dilatation and bile duct invasion. The reasons might lie not only in the fact that bile duct dilatation may be more often reflected in bile duct compression in our small sample study, but also in the aforementioned problem that pathological and imaging features do not correspond in one-to-one correspondence due to retrospective studies. In addition, bile duct dilatation was not an independent predictor of long-term prognosis in our study. The potential reason for this may be that no bile duct invasion was found in any of the lesions in our study where bile duct dilatation was present.

Based on the results of our study, we found that peritumoral hypointensity on HBP had no prognostic predictive effect, nor did we find it to be associated with pathologic features. Peritumoral hypointensity on HBP indicates that gadoxetic acid uptake in nontumorous liver parenchyma is reduced by decreased or dysfunctional hepatocytes due to arterioportal shunts, portal vein obstruction, bile duct obstruction, microvascular invasion of hepatocellular carcinoma, sinusoidal obstruction, focal eosinophilic infiltration, peliosis, fibrosis, inflammation, or any combination of these [36]. As many factors are capable of influencing peritumoral hypointensity on HBP, we have not found a correlation with pathologic features. What is more, we also found no association between peritumoral hypointensity on HBP and prognosis because not all of these factors were associated with prognosis.

There are some limitations in the present study. First, due to the exclusion of patients without preoperative gadoxetic acid-enhanced MRI and those who received preoperative treatment, the number of patients was small. Therefore, our results are supportive but not conclusive. Second, it is a retrospective analysis, which may have suffered from bias in selecting operative procedures and lack of accurate correspondence between imaging features and pathological features. In the future, further validation should be done through multicenter and prospective studies. Lastly, the number of patients with peritumoral early enhancement was limited in both groups, although the two groups showed significant differences in LRFS and RFS. Our data are not definitive but represent a preliminary result and further external validations are needed.

In conclusion, based on the PSH principle, AR has the potential to improve LRFS in patients with CRLM, although it did not show any improvement on RFS and OS. Meanwhile, peritumoral early enhancement with CRLM indicated poor LRFS and RFS in patients who had undergone NAR. Therefore, peritumoral early enhancement can be used as a reference indicator in the choice of surgical procedure and could further decrease the risk of postoperative recurrence.