Introduction

Colorectal cancer is common and its incidence is increasing [1]. Fifty percent of all patients with colorectal cancer develop metastasis, and colorectal liver metastasis (CRLM) is the most frequent [2]. Liver resection is the most effective therapy for CRLM, and the 5-year survival rate is as high as 50% [3]. However, postoperative recurrence is common and remains an important issue to be addressed. Previously, several prognostic factors of recurrence were reported including carcinoembryonic antigen (CEA), metastatic number, tumor size or lymph node metastasis [4, 5]. In these previous reports, prognostic models have been described based on calculus while using these prognostic factors; however, these models have not been broadly adopted in the clinical settings. Therefore, to improve patient prognosis, we must identify simple prognostic markers in CRLM.

D-dimer, a fibrin cleavage product, is a clinically important marker used to diagnose pulmonary thromboembolism [6]. Recently, D-dimer levels were reported to be strongly related to various types of colorectal cancers [7, 8]. In addition, D-dimer levels show correlation with tumor progression, recurrence, and prognosis in other cancers (lung cancer [9, 10], prostate cancer [11], and gastric cancer [12]). However, there have been no reports evaluating the correlation between D-dimer levels and CRLM recurrence.

Therefore, we aimed to evaluate D-dimer levels in patients with CRLM to predict the prognosis of CRLM, following liver resection. We also evaluated the association between D-dimer levels and recurrence, clinicopathological factors, inflammatory factors, and patient survival in CRLM.

Materials and methods

Patients and samples

We retrospectively evaluated 90 consecutive patients with liver metastasis of colorectal cancer who underwent curative liver resection at our department from 2006 to 2017. The patients were aged 38–83 years. The various blood examinations (including D-dimer) were measured before a month of operation. The surgical techniques applied were partial resection (46 cases), subsegmentectomy (14 cases), segmentectomy (17 cases), and lobectomy (13 cases). The locations of the primary cancer were, the cecum (6 cases), ascending colon (6 cases), transverse colon (4 cases), descending colon (4 cases), sigmoid colon (14 cases), and rectum (26 cases). Tumor staging was performed based on the criteria of the Union for International Cancer Control tumor-node-metastasis staging system (seventh edition). The recurrence was observed in 51 cases, and the location of the metastasis were, liver (19 cases), lung (22 cases), peritoneum (4 cases), lymph node (12 cases), colon local site (2 case), and others [brain (2 cases), bone (2 cases), abdominal wall (1 case), pleura (1 case)]. The adjuvant chemotherapy was performed to 69 cases as follows: FOLFOX (5 cases), bevacizumab plus FOLFOX (1 cases), Capecitabine (9 cases), XELOX (18 cases), bevacizumab plus XELOX (3 cases), FOLFIRI (3 cases), cetuximab plus FOLFIRI (2 cases), bevacizumab plus SOX (1 case), TS-1 (4 cases), UFT (8 cases), UFT/UZEL (6 cases) and 5-FU/LV (1 case). All patients provided written informed consent for the treatment and data analyses, while the study design was approved by our Clinical Ethics Committee (http://ciru.dept.showa.gunma-u.ac.jp/guidance/storage-sample/list.html).

Statistical analysis

All statistical analyses were performed with EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria) [13]. Level of significance was evaluated by performing Student’s t test, analysis of variance, Mann–Whitney U test, and by determining the Pearson’s correlation coefficient. Survival curves were created by applying the Kaplan–Meier method and analyzed by performing the log-rank test. Prognostic factors were examined in univariate and multivariate analyses, using a Cox proportional hazards model. The cut-off value for D-dimer was evaluated using the receiver operating characteristic (ROC) curve analysis. Differences were considered significant at P < 0.05.

Results

D-dimer levels by postoperative recurrence groups

The average preoperative D-dimer level was 1.21 µg/mL (range 0–9.3 µg/mL). Compared to the no-recurrence group, the recurrence group had significantly higher D-dimer levels (P = 0.00736*; Fig. 1a). Other coagulation markers (FDP, platelet counts, PT and APTT) had no correlation to recurrence. The optimal cut-off value for D-dimer levels was 0.6 µg/mL [area under the curve (AUC) 0.664, sensitivity 0.683, specificity 0.714; Fig. 1b]. This cut-off value was used for the following analyses. Two typical cases (one is a low D-dimer case and the other is a high D-dimer case) are shown in the Supplementary Figure.

Fig. 1
figure 1

Significant difference occurred in the recurrence group with higher D-dimer levels than in the no-recurrence group (P = 0.00736; a). The optimal cut-off value for D-dimer levels (0.6 µg/mL) was determined using the receiver operating characteristic curve analysis [area under the curve (AUC) 0.664, sensitivity 0.683, specificity 0.714; b]

Full size image

Clinicopathological factors by D-dimer groups

The significant differences in various clinicopathological factors with D-dimer levels are shown in Table 1. Significant differences occurred in various clinicopathological factors with the high D-dimer group (≥ 0.6 µg/mL) including CEA (P = 0.0027*) and synchronous metastasis (P = 0.022*). A borderline significant difference was also found in primary tumor location (left-sided colon) with the high D-dimer group (P = 0.0655). In contrast, liver metastasis factor, tumor size, tumor number and venous thrombosis (VTE) showed no significant differences by D-dimer levels. Moreover, no significant differences were also observed according to D-dimer levels in inflammatory factors, the neutrophil/lymphocyte ratio (NLR), and modified Glasgow Prognostic Score (mGPS). In addition, D-dimer had negative correlations with albumin (R2 − 0.434, P = 0.0000294*; Fig. 2a) and size of liver metastases (R2 0.217, P = 0.0401*; Fig. 2b).

Table 1 D-dimer and clinicopathological factors
Full size table
Fig. 2
figure 2

D-dimer had negative correlations with albumin (R2 − 0.434, P = 0.0000294*; a) and the size of liver metastases (R2 0.217, P = 0.0401*; b)

Full size image

Associations between D-dimer levels and postoperative recurrence and prognosis

The ability of D-dimer levels to predict postoperative recurrence-free survival (RFS) and cancer-specific survival (CSS) is shown in Fig. 3. Compared to the low D-dimer group, the high D-dimer group showed poorer RFS (P = 0.0000841*; Fig. 3a) and poorer CSS (P = 0.00615*; Fig. 3b). D-dimer levels exhibited a stronger correlation with RFS (Table 2) and CSS (Table 3) after liver resection for CRLM. D-dimer predicted recurrence in the univariate and multivariate analyses [hazard ratio (HR) 3.08, 95% confidence interval (CI) 1.52–6.67, P = 0.00179*; Table 2]. Furthermore, as shown in Table 2, D-dimer provided greater predictive value than CEA (HR 1.81, 95% CI 0.934–3.51, P = 0.0784), primary T factor (HR 1.83, 95% CI 0.999–3.39, P = 0.0539), and adjuvant therapy (HR 0.400, 95% CI 0.195–0.822, P = 0.0127*). In Table 3, D-dimer shows correlation with CSS in the univariate analysis and tends to relate with CSS in the multivariate analysis (HR 2.51, 95% CI 0.879–7.15, P = 0.0857*; Table 3).

Fig. 3
figure 3

Relationship between postoperative survival and D-dimer levels. Kaplan–Meier curves for the low and high D-dimer groups are shown. a High D-dimer levels were associated with poor recurrence-free survival (RFS; P = 0.0000841) and b high D-dimer levels were also associated with poor cancer-specific survival (CSS; P = 0.00615)

Full size image
Table 2 Univariate and multivariate analysis of prognostic factors for RFS using Cox proportional hazards model
Full size table
Table 3 Univariate and multivariate analysis of prognostic factors for CSS using Cox proportional hazards model
Full size table

Associations between D-dimer levels and adjuvant chemotherapy on postoperative recurrence and prognosis

The correlation between postoperative prognosis and adjuvant chemotherapy was also analyzed. The adjuvant chemotherapy group showed improvement in RFS (P = 0.00203*; Fig. 4a) but had no correlation to CSS (P = 0.318; Fig. 4b). Moreover, in the D-dimer low group (less than 0.6 µg/mL), adjuvant chemotherapy led to improvement in DFS (P = 0.018*; Fig. 4c) and did not relate to CSS (P = 0.249; Fig. 4d). In the D-dimer high group (above 0.6 µg/mL), adjuvant therapy had no correlation to both DFS (P = 0.0618; Fig. 4e) and CSS (P = 0.857; Fig. 4f).

Fig. 4
figure 4

Relationship between postoperative survival and adjuvant chemotherapy. Adjuvant chemotherapy group showed improvement in RFS (P = 0.00203*; a) but had no correlation to CSS (P = 0.318; b). In addition, Kaplan–Meier curves for adjuvant chemotherapy are shown in each D-dimer low group (less than 0.6 µg/mL) and D-dimer high group (above 0.6 µg/mL). In the D-dimer low group, adjuvant chemotherapy was associated with improving poor recurrence-free survival (RFS; P = 0.018; c) but was not associated with cancer-specific survival (CSS; P = 0.249; d). In the D-dimer high group, adjuvant chemotherapy was not associated with RFS (P = 0.0618; e) and CSS (P = 0.857; f)

Full size image

Discussion

In the present study, we showed that high D-dimer levels (≥ 0.6 µg/mL) were associated with postoperative recurrence after hepatic resection for CRLM. Thus, D-dimer levels might be a clinically useful prognostic marker for CRLM.

Zacharski et al. reported that cancer cells promote coagulation by inactivating the fibrinolytic system (e.g., via urokinase-type plasminogen activator); and different carcinoma types may promote coagulation via numerous mechanisms [14]. Gene et al. also reported that breast and prostate cancers were associated with elevated levels of coagulation factors (e.g., vascular endothelial growth factor, interleukin-6, soluble P-selectin, fibrinogen, and D-dimer) [15]. Thus, cancer cells appear to activate coagulation. Moreover, Mytnik et al. reported that colorectal cancer activates several coagulation factors, including plasminogen activator inhibitor-1, prothrombin fragments, and protein C [16]. The correlation between colorectal cancer and D-dimer was first reported by Edwards et al. in 1993 [17], and the initial finding was supported by the findings of subsequent studies. In colorectal cancer, D-dimer was shown to be a prognostic factor after the resection of the primary lesion [7, 18,19,20,21], and an indicator of therapeutic effect of chemotherapy [22].

In previous studies, several prognostic factors of CRLM were reported, including metastatic number, differentiation, resected margin, and the tumor marker, CEA [1, 2]. Moreover, several studies have proposed the use of calculus or nomogram to determine prognosis based on the above risk factors [4, 5, 23]. However, the relationship between D-dimer and CRLM had not been reported previously. In the present study, high D-dimer level correlated with poorer RFS and CSS among patients with CRLM and was an independent prognostic factor of RFS in multivariate analysis. In addition, D-dimer level was a more significant predictor than CEA, primary T factor, or adjuvant therapy. Blackwell et al. also reported that D-dimer reflected the prognosis of metastatic colorectal carcinoma more than CEA [20]. Therefore, D-dimer levels could conveniently predict the prognosis in CRLM.

Nevertheless, its mechanism is still unclear. Recent reports have shown high D-dimer levels are related to poor prognosis, probably based on chemoresistance in various cancers, and/or circulating tumor cells (CTC). Inanc M et al. reported D-dimer was a marker of chemosensitivity [24] and Zhu L et al. reported high D-dimer correlates to poor prognosis when treated with bevacizumab [25]. Tomimaru Y et al. proved D-dimer could be used as a predictor for neoadjuvant chemosensitivity in esophageal cancer by evaluating pathological response [26]. In our study, low D-dimer cases were shown to improve DFS by adjuvant chemotherapy, but high D-dimer cases had no correlation prognosis and adjuvant chemotherapy. Tissue factors play a role in chemoresistance thorough hypercoagulopathy state, and a high level of VEGF was also predictive of chemoresistance on breast cancer [27]. From these things, D-dimer reflected hypercoagulopathy state on cancer patient and as a result, high D-dimer level lead to chemoresistance.

Among the clinicopathological factors examined, significant differences occurred with D-dimer in CEA (P = 0.0027*) and synchronous metastasis; therefore, D-dimer levels may reflect cancer progression. CEA has been reported in several studies as a prognostic factor in colorectal cancer, and found to be correlated with CRLM prognosis [28, 29]. Recently, right-sided colon cancer was reported to have a poor prognosis [30] and different molecular biological background (KRAS/BRAF mutation, and p53) compared with left-sided cancer [31]. In this study, right-sided cancer in the primary anatomical site showed tendency to have difference with high D-dimer level (P = 0.0655).

The circulating tumor cells (CTCs) are proved to relate to tumor progression and prognosis in various cancers. Recently, it is reported CTCs had relation to hypercoagulability and D-dimer could be excellent predicator [11, 32]. In colorectal cancer, CTC positive cases had poor prognosis after liver resection to CRLM [33, 34]. In our study, D-dimer may indicate poor prognosis or hypercoagulability as a consequence of reflecting CTCs behavior.

This study has several limitations. Because it is a retrospective study, we could not evaluate D-dimer level on postoperative state. In addition, it is unclear how tumor progression is mechanistically related to D-dimer levels. Therefore, further prospective studies are needed to evaluate the underlying mechanism.

Recently, the antitumor effect of anticoagulant was reported in an investigation aimed at determining the correlation between cancer and hypercoagulability [35]. In addition, Klerk et al. reported that low molecular weight heparin increased survival time for solid tumor patients [36]. In small cell lung cancer, the combination of chemotherapy and heparin led to improved antitumor effect [37]. Moreover, according to Rothwell et al., long-term administration of aspirin reduced the risk of colorectal cancer development [38]. These findings, regarding coagulation and cancer, indicate coagulative factors as promising therapeutic targets and D-dimer as a predictive marker of therapeutic response in cancer.

Although liver resection greatly improves the prognosis of CRLM, it is of importance to prevent recurrence after resection. Therefore, we sought to find clinically useful factors for predicting recurrence after the resection of CRLM. Moreover, our data showed D-dimer high cases did not have enough effect of adjuvant chemotherapy as compared to D-dimer low groups. The anticoagulant therapy may be new target instead of chemotherapy for high D-dimer cases.

In conclusion, our study provides the first evidence that D-dimer may be a novel biomarker of prognosis in CRLM in various aspects. We showed that high D-dimer level was associated with poor RFS and CSS in patients with CRLM. Therefore, D-dimer may be a useful biomarker of recurrence and prognosis in CRLM. Our findings also indicate that coagulation factors may be promising targets for controlling cancer progression.