Introduction

Preoperative rectal cancer staging is based on two key principles, namely, prognostic staging and defining the anatomy in order to plan treatment. Currently, many studies on preoperative MRI have concentrated on T and N staging and the circumferential resection margin (CRM). The application of multimodal neoadjuvant treatment on the basis of such MRI findings has markedly decreased local recurrence rates. However, distant metastasis (including synchronous and metachronous distant metastases) remains a problem in patients with rectal cancer [1].

Clinicopathological factors including TNM staging, lymphovascular invasion (LVI), histological tumour grade and pretreatment serum level of carcinoembryonic antigen (CEA) are widely accepted prognostic factors of rectal cancer [2–4]. A recent published study revealed that positive extramural vascular invasion, high T stage and positive regional lymph node metastasis are significantly associated with a high risk of synchronous distant metastasis [4].

The advent of quantitative MR imaging has facilitated an increased role for imaging in risk stratification. Specifically, dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in combination with pharmacokinetic modeling of the contrast agent distribution in malignant tissue can be used to assess tumour microcirculation in vivo and noninvasively [5, 6]. It assesses the extravasation of paramagnetic contrast agents such as gadolinium-chelates and can be applied to organs with little motion such as the rectum [7]. However, to our knowledge, the actual clinical benefit of DCE-MRI to predict synchronous distant metastasis and the prognostic aspects in distant metastatic rectal cancer has not yet been well determined.

Therefore, the purpose of this retrospective study was to explore the correlations between DCE-MRI quantitative parameters and synchronous distant metastasis and the clinicopathological factors in rectal cancers.

Material and methods

Patients

This retrospective study was approved by our institutional review board and the requirement for informed consent was waived. Between September 2013 and December 2014, patients with pathologically confirmed rectal adenocarcinoma at our hospital were screened. Patients who underwent surgery alone after MRI examination were included in the study. Those patients were initially investigated via MRI of the rectum and a CT examination for both local and distant staging, and also followed up with a CT examination of the chest, abdomen and pelvis at 6 months post-diagnosis. Patients with the following criteria were excluded: (a) patients receiving any neoadjuvant treatment; (b) DCE-MRI not included in the MRI of the rectum; (c) patients who were lost to follow-up; or (d) images showing unacceptable distortion due to artefacts from metallic clips around the tumours.

MRI examination

All MRI examinations were performed using a 3.0-Tesla MRI scanner (MAGNETOM TrioTim, Siemens, Erlangen, Germany) with a 16-element pelvic phased-array coil. Non-enhanced MR imaging, including oblique axial T1-weighted 2D turbo spin-echo images (repetition time ms/echo time ms, 722/11; section thickness, 3.0 mm; intersection gap, 0.3 mm; field of view, 25 × 25 cm; and matrix, 384 × 326) and sagittal and oblique axial and oblique coronal T2-weighted 2D turbo spin-echo images (repetition time ms/echo time ms, 4,000/99; section thickness, 3.0 mm; intersection gap, 0.3 mm; field of view, 25 × 25 cm; and matrix, 384 × 326) were performed. Three pre-contrast Vibe 3D gradient-echo images (TR/TE 5.32/1.81, slice thickness, 3.0 mm, field of view, 25 × 25 cm, matrix 256 *161, 36 axial slices) with three different flip angles (5°, 10° and 15°) were acquired for the purpose of T1 mapping. Thereafter the DCE images (repetition time ms/echo time ms, 5.32/1.81; section thickness, 3.0 mm; intersection gap, 0.3 mm; field of view, 25 × 25 cm; and matrix, 256 × 161; number of phases, 25; temporal resolution,12.2 s; and the total imaging time, 5 min 5 s) were obtained. An automated injector system (Stellant MR Injection System, Medrad, Germany) was used to apply the Omniscan (GE HealthCare, 0.5 mmol/ml; Dosage (ml) = weight (kg) × 0.2 ml/kg) by bolus injection with a flow rate of 2.5 ml/s. MRI total acquisition time was around 30 min.

Imaging analysis

Two experienced radiologists with 10 and 4 years of experience, respectively, in abdominal imaging evaluated the data independently, blinded to all clinical information and the reference standard. For DCE-MRI parameter measurements, pharmacokinetic analysis was carried out using non-commercial software, Omnikinetics (GE Healthcare, Shanghai, China) with the two-compartment Extended Tofts model. A fully automated image-based individualised AIF (iAIF) estimation method was used as the AIF estimation method. The parameters including the volume transfer constant between blood plasma and extracellular-extravascular space (EES) (K trans, min-1), rate constant between EES and blood plasma (K ep , min-1) and volume of EES space per unit volume of tissue (V e ) were derived from the entire ROI for each patient. The radiologists manually drew ROIs along contours of the tumour, covering the whole lesion with the exclusion of the necrosis, peripheral fat and blood vessels based on T2-weighted images as a guide. After each ROI placement, colour-coded, pixel-wise, parametric perfusion maps of K trans, K ep and V e were generated with sample patients as shown in Fig. 1 [8].

Fig. 1
figure 1

A 73-year-old man with rectal carcinoma who had liver metastasis at initial staging. (a) On T-2-weighted turbo spin-echo (TSE) axial image, a mass (arrow) was seen. The colour-coded K trans map (b), K ep map (c) and V e map (d) shows mixed red, green and blue colours in the corresponding tumour with a K trans of 0.187 × 10−3 min-1, K ep 1.412 min-1 of and V e of 0.145

Full size image

Clinicopathological examinations and follow-up

Clinicopathological prognostic factors were derived from the patients’ records including age, plasma carcinoembryonic antigen (CEA) level, T and N stage, and tumour differentiation grade. Pathological T and N stage, tumour differentiation grade and the presence of microscopic LVI were confirmed by a pathologist with 6 years of experience in pathology. The lymph node stage was categorized into negative (no metastasis in regional nodes), positive (including N1: metastasis in 1–3 nodes and N2: metastasis in four or more nodes). Tumour differentiation grade was categorized into three stages (well, moderately and poorly differentiated).

The lung and liver metastasis lesions that were pathologically confirmed either underwent surgery resection or biopsy. Among patients in whom the metastasis lesions were not pathologically confirmed, contrast-enhanced liver MRI or fludeoxyglucose–positron emission tomography/computed tomography (FDG–PET/CT) are used for more precise evaluation. A clinical follow-up of 6 months was undertaken to confirm the final results of distant metastases. Newly detected metastatic lesions within 6 months after the initial diagnosis and those with distant metastases at initial staging were considered synchronous metastasis.

Statistical analysis

Statistical analyses were performed using software (SPSS Statistics for Windows, version 20.0, IBM Corp, Armonk, NY, USA; MedCalc, version 11.5, Mariakerke, Belgium; and R, version 3.1.3). To test the reproducibility of DCE-MRI measurements independently obtained by two observers, intra-class correlation coefficient (ICC) values for the parameters of two DCE-MRI data sets were calculated with a 95 % confidence interval (CI). Differences in DCE-MRI parameters with regard to clinicopathological factors were tested. To compare the variables, the normality and homoscedasticity of the multiparametric data were tested using the Q-Q plots and the Levene tests. Data subjected to normal distribution were subjected to Student’s t test or analysis of variance (ANOVA). Conversely, data not subjected to normal distribution were analysed using the Mann-Whitney U test or Kruskal-Wallis H(K) test. Receiver operating characteristic (ROC) analyses were used to find a reasonable threshold of K trans, K ep and V e to differentiate lesions with synchronous distant metastasis from those without metastasis. For the ROC analyses, the area under the curve (AUC) was calculated and AUCs were compared using nonparametric methods [9] and optimal thresholds were obtained maximising the Youden index [10]. A p value less than 0.05 was considered statistically significant.

Results

Patients

Table 1 summarizes the clinicopathological characteristics. Among 63 enrolled patients, 31 patients were confirmed to have synchronous distant metastases. Among the 31 patients in the synchronous distant metastasis group, 21 were men and ten were women. The mean age was 58.5 years (range: 34–77). The mean tumour length was 4.84 cm (range: 2.32–7.9). The mean age in the non-metastasis group (n = 32; 18 men and 14 women) were 62.61 years (range: 35–78), respectively. The mean tumour length of the non-metastasis group was 4.36 cm (range: 0.8–7.46 cm).

Table 1 Summary of clinicopathological characteristics of included lesions
Full size table

Among the 31 patients in the synchronous distant metastasis group, 26 patients had distant metastases at the time of initial diagnosis and five patients showed newly developed liver metastases within 6 months of surgery. Hepatic metastases were the most common metastases (19 patients, 61.3 %) followed by lung metastases (nine patients, 29.0 %). Three patients (9.7 %) had both liver and lung metastases; distant lymph node metastases and adrenal gland metastases was not observed. Distant metastases were confirmed pathologically by surgery in 18 patients and by sequential radiological follow-up in 13 patients. Among them, five patients underwent FDG–PET/CT, five patients underwent hepatocyte-specific contrast-enhanced liver MRI and three patients underwent both PET/CT and hepatocyte-specific contrast-enhanced liver MRI for more precise evaluation.

The median time interval from MRI to surgery was 7 days (range: 3–27). Patients in the metastasis group received low anterior resection (LAR, n = 21), abdominoperineal resection (APR, n = 8) and extended resection (n = 2). Patients in the non-metastasis group received LAR (n = 19), APR (n = 12) and extended resection (n = 1).

DCE-MRI parameters

Comparison of the DCE-MRI parameters according clinicopathological status is shown in Table 2.

Table 2 Comparison of the DCE-MRI parameters according to clinicopathological status
Full size table

Higher K trans (greater flow and permeability) values from whole transverse ROIs obtained by DCE-MRI were noted among patients with synchronous distant metastasis than those without distant metastasis (p < 0.001). The k ep and V e were also significantly higher in the distant metastases group (p = 0.022 and 0.034, respectively). Figure 2 illustrates the distribution of K trans, K ep and V e values in the synchronous distant metastasis and non-metastasis group.

Fig. 2
figure 2

Boxplots showing K trans, K ep and V e in lesions with and without distant metastasis. K trans (a), K ep (b) and V e (c) were significantly higher in the lesions with distant metastasis than in the lesions without distant metastasis (0.536 ± 0.242 vs. 0.299 ± 0.118 min-1, p < 0.001; 1.598 ± 0.477 vs. 1.341 ± 0.390 min-1, p = 0.022; and 0.324 ± 0.173 vs. 0.249 ± 0.091, p = 0.034, respectively)

Full size image

The ROC analysis was used to find the reasonable threshold of K trans, K ep and V e to differentiate lesions with synchronous distant metastasis from those without metastasis (Table 3). The K trans showed the highest AUCs with 0.788 (p < 0.001), with sensitivity of 61.29 % and specificity of 87.5 %, respectively. K trans showed a significantly higher AUC (0.788 vs. 0.645) than V e (p = 0.008). K trans and K ep showed a fair degree of accuracy (p = 0.165), as did K ep and V e (AUC: K ep 0.689, V e 0.645; p = 0.674).

Table 3 Effectiveness of DCE-MRI parameters for discriminating lesions with synchronous distant metastasis from those without metastasis
Full size table

The K trans, K ep and V e values of the pLVI positive group were significantly higher than those of the pLVI negative group (p < 0.001, p = 0.016 and 0.004; respectively) (Fig. 3).

Fig. 3
figure 3

Boxplots showing K trans, K ep and V e in the pLVI-positive and -negative groups. K trans (a), K ep (b) and V e (c) were significantly higher in the pLVI-positive group than the pLVI-negative group (0.562 ± 0.271 vs. 0.343 ± 0.151, p < 0.001; 1.659 ± 0.580 vs .1.372 ± 0.339, p = 0.016; and 0.358 ± 0.204 vs. 0.250 ± 0.078, p = 0.004, respectively)

Full size image

The K trans and V e correlate with tumour histological grade, the K trans values increased from 0.193 (min-1, well differentiated) to 0.478 (min-1, poorly differentiated). The V e values increased from 0.171 (well differentiated) to 0.326 (poorly differentiated). In addition, a statistically significantly higher V e value was found in CEA ≥ 5 ng/ml group (p = 0.04).

No statistically significant correlations were found for K trans, K ep and V e values in tumours of different histological T and N stage.

In the evaluation of inter-observer agreement for the independent sets of DCE-MRI parameters, ICC values for each parameter showed an almost perfect level of agreement; 0.889 (95 % CI 0.829–0.950) for K trans, 0.862 (95 % CI 0.788–0.937) for K ep and 0.721 (95 % CI 0.581–0.861) for V e .

Discussion

In this study, we evaluated the usefulness of perfusion parameters derived from DCE-MRI for assessing the correlations with synchronous distant metastasis and the clinicopathological factors in rectal cancers. Our results showed that tumours with synchronous distant metastasis had higher K trans and K ep values. Angiogenesis is one of the fundamental factors in the formation of distant metastases [11]. In terms of biological explanations, we would propose that tumours with higher vascular permeability are more adherent to distant metastasis. DCE-MRI parameters K trans correlate with vascular permeability, and hence angiogenesis within tumour tissue. Similarly, higher K ep value represents greater blood return to vasculature. In addition, K ep is only affected by the contrast concentration and fractional volumes in the tumour EES and might thus more accurately reflect the tumour capillary permeability [12]. It is reported that DCE-MRI parameters permit the noninvasive estimation of tumour angiogenesis [5, 13], including rectal cancer [14]. Thus, the correlation between the imaging data and the distant metastases was not unexpected. These results are in line with the findings of other authors [6, 15].

In addition, with respect to the contrast agent exchange constant V e , in our study, V e was significantly higher in synchronous distant metastasis group. It has been shown that the loss of function of cell–cell adhesion molecules, such as e-cadherin, is a crucial step in tumour progression, both regional and metastatic [16–18]. This loss of function leads to a larger interstitial space, which is reflected by a higher V e on DCE-MRI. Our results were in agreement with those of a previous study in glioma [19, 20].

A preoperative CEA level ≥ 5 ng/ml was reported to be associated with the early distant metastasis [21, 22]. Villeneuve et al. [23] suggested that preoperative CEA level ≥ 10 ng/ml was prognostic factor in patients with lung metastasis from colorectal cancer. In our study, significant correlation was also found between CEA level and synchronous distant metastasis (data not show). In addition, a statistically significant higher V e value was found in CEA ≥ 5 ng/ml group (p = 0.04).

Regarding the Davenport et al. [24] investigation, V e seems to be a more robust and reproducible biomarker. However, our findings revealed lower accuracy of V e than K trans to access synchronous distant metastasis. Further research about the correlation between V e and synchronous distant metastasis is deemed necessary.

In our study, correlations were found between the K trans, K ep and V e values and pLVI status. Recent studies demonstrate that there is a significant difference in the incidence of metastatic disease between those with and those without mesorectal vascular invasion in rectal carcinoma [25, 26]. Venous invasion is an independent risk factor for local and distant recurrence and poorer overall survival [27, 28].

We used lymphovascular invasion instead of vascular invasion analysis on the basis of routine staining in this study. However, histopathological LVI may be a different entity from vascular invasion. Histologically, lymphatic vessels and blood vessels are both thin walled and tubular structures surrounded by endothelial cells and it is often not possible to accurately differentiate them without elastic tissue staining. As a result, the term ‘lymphovascular’ is usually used [29]. As the DCE-MRI parameters mainly characterize angiogenesis rather than lymphangiogenesis, the correlation between the imaging data and the vascular invasion should be assessed in future work. MRI-detected extramural vascular invasion (EMVI) is an independent risk factor for distant metastasis in rectal cancer [25, 26]. The comparison of the ability between DCE-MRI parameters and EMVI to predict distant metastasis was not addressed in this study.

The results of this study also imply that perfusion parameters may serve as an effective image-based indicator in the differentiation of tumour histological grade. The poorly differentiated rectal cancer typically presents with hypervascularity accompanied by increased permeability and leakiness. The MR contrast agents (in the present study, the Omniscan) that leak slowly through the intact basement membrane are able to escape more quickly through tumour vessels into the EES.

We are now in an era in which treatment is tailored according to individual risk. Early distant metastasis is associated with poor survival [21]. Control of distant metastasis is presently a major concern for the treatment of locally advanced rectal cancer [1]. Preoperative identification of distant metastasis is especially important for some reasons: first, patients with synchronous metastasis should undergo different treatment, such as neoadjuvant chemotherapy or metastasectomy. Neoadjuvant chemotherapy with a full systemic dose could have the potential to treat micrometastases before they can progress to more advanced and detectable metastases [30]. Some clinical trials including preoperative induction chemotherapy for intense systemic control before standard treatment have been performed [31]. The main targets of this type of clinical trial are patients with the greatest risk of distant metastasis. Our results indicate that DCE-MRI might be helpful to select candidates for associated clinical trials. Second, patients with high-risk findings on MRI could undergo liver-specific contrast-enhanced MRI or FDG–PET to detect hidden metastases. It has been shown that gadoxetic-acid-enhanced MRI at 3-T is significantly more sensitive than 64-row multidetector CT for the detection of small (<10 mm) hepatic metastases [32]. As we found a correlation between DCE-MRI data and the occurrence of synchronous distant metastases, further prospective investigations with more patients and a long-term follow-up are needed to assess the potential of DCE-MRI to predict survival.

There are some limitations in this study. First, we have proven the value of detection of synchronous distant metastasis; however, metachronous metastasis remains an uncertain area. Although prediction of synchronous metastasis has great value for treatment planning, the relapse-free survival of patients with rectal cancer is directly related to metachronous metastasis, rather than synchronous metastasis. Second, in this study, we chose patients who underwent surgery alone. Patients who received any neoadjuvant treatment were excluded for the following reasons: pathological results like tumour T staging, N staging, tumour differentiation grade and lymphovascular invasion may change after the preoperative chemoradiotherapy. Preoperative chemoradiotherapy may induce tumour downstaging or even pathological complete response (pCR) in some patients. Neoadjuvant treatment may influence the results of distant metastasis. This may predispose to selection bias. Third, the relatively small number of subjects and low number of patients with a subgroup of histological T1 stage and histologically well-differentiated grade may limit the applicability of results.

In conclusion, these findings in our study suggest that DCE-MRI data may represent a prognostic indicator for synchronous distant metastases in patients with rectal cancer. We believe that perfusion parameters derived from DCE-MRI might help to identify these high-risk patients before therapy.