Late gadolinium enhancement of colorectal liver metastases post-chemotherapy is associated with tumour fibrosis and overall survival post-hepatectomy

Colorectal cancer is the second leading cause of cancer deaths in the developed world [1]. Approximately half of patients develop liver metastases and most deaths are related to metastatic disease [2]. The median survival of patients with colorectal liver metastases (CRCLM) without treatment is 7.5 months [3]. With advancements in surgical and chemotherapy techniques, the survival of patients with CRCLM has significantly improved. In a meta-analysis by Kanas et al. (2012), the 5- and 10-year survival of patients with resected CRCLM was 38% and 26%, respectively [2]. This is likely even higher with more recent data and improving surgical and chemotherapy techniques.

The ability to predict prognosis informs treatment recommendations, including surgery and/or chemotherapy. Several prognostic indicators stratify risk for patients with CRCLM including clinical, pathology, and molecular prognostic biomarkers (3-4). However, the use of magnetic resonance imaging (MRI) to stratify risk in patients with CRCLM is relatively unexplored. MRI is routinely used clinically for diagnosis, staging, and operative planning in patients being considered for liver resection, so information gained from MRI could be easily translated into clinical practice.

Several studies have demonstrated that tumour fibrosis in post-hepatectomy CRCLM specimens is associated with overall survival [4, 5]. This may be related to the pathological response to chemotherapy. Pathologically, tumour fibrosis in CRCLM closely resembles the appearance of tumour fibrosis in cholangiocarcinoma. In cholangiocarcinoma, late gadolinium enhancement on MRI with extracellular contrast agents is correlated with tumour fibrosis [6]. This association has also been reported with CRCLM, although this is less well studied [6]. Thus, we hypothesise that late gadolinium enhancement of colorectal cancer liver metastases may be correlated with tumour fibrosis post-chemotherapy and therefore with overall survival post-hepatectomy.

The purpose of our study was to determine whether late gadolinium enhancement of CRCLM on MRI with gadobutrol is associated with tumour fibrosis and overall survival post-hepatectomy.

This study was an institutional-REB approved, retrospective study.

Among the 121 patients who met inclusion/exclusion criteria for the study (Fig. 2), the mean age was 63.0 years (SD: 11.2 years) with 70 (57.9%) males and 51 (42.1%) females (Table 1). The median time from MRI to surgery was 2.7 months (range: 0.1-10.5 months). There were 40 deaths during the follow-up period.

Based on the Youden Index, the optimal cut-off for weak and strong TTE was CNR = 11. Seventy-four patients (61.1%) had weak and 47 patients (38.8%) strong TTE.

Patients with strong TTE were more likely to have smaller tumours (p = 0.019). No other demographic data were significantly different between the MRI groups (Table 1).

It was possible to accurately match lesions between MRI and pathology for 91 patients with 126 target CRCLMs. The Spearman correlations between TTE and the mean target percentage of fibrosis, necrosis, acellular mucin, and viable tumour cells were 0.43 (p < 0.001), -0.22 (p = 0.036), 0.02 (p = 0.84), and -0.05 (p = 0.63), respectively. The median target percentage of fibrosis for high TTE and low TTE were 15.0% [interquartile range (IQR): 3.0% to 30.0%] and 37.5% (IQR: 15.0% to 51.3%), respectively (Fig. 3a). The median target percentage necrosis for high TTE and low TTE was 30.0% (IQR: 15.0% to 50.0%) and 10.0% (IQR: 3.8% to 35.0%), respectively (Fig. 3b). The median target percentage viable tumour cells for high and low TTE was 40.0% (IQR: 10.0% to 50.0%) and 32.5% (IQR: 16.5% to 50.0%) respectively (Fig. 3c). Most patients did not have tumours that contained acellular mucin (only 12 patients); therefore, the median target percentage acellular mucin was 0% for both high and low TTE.

Strong TTE was associated with survival on univariate analysis (p = 0.003). At 3 years, 88.4% of patients with strong TTE on the preoperative MRI were alive vs. 58.8% in patients with weak TTE (Fig. 4).

One hundred twelve patients (with 34 events) had complete data available for the multivariable analysis. TTE had an adjusted hazard ratio of 0.32 (95% CI: 0.14-0.75, p = 0.008). The adjusted hazard ratio of the clinical risk score was 2.41 (95% CI: 1.19-4.90) (Table 2).

Post-hoc sensitivity analyses were performed for time from MRI to surgery and tumour size ≥ 5 cm. None of these variables were found to be significant contributing variables on our sensitivity analysis. When time from MRI to surgery was included in the Cox regression model, TTE had an adjusted hazard ratio of 0.33 (95% CI: 0.14-0.75, p = 0.009). When tumour size was included in the Cox regression model, TTE had an adjusted hazard ratio of 0.33 (95% CI: 0.14 to 0.77, p = 0.010).

Inter-rater reliability was very good with a relative intraclass correlation of 0.84 (95% CI: 0.77-0.89).

In this study, we demonstrated that late gadolinium enhancement of CRCLM post-chemotherapy on gadobutrol-enhanced MRI post-hepatectomy is associated with tumour fibrosis and with overall survival, after taking into account known clinical prognostic factors. The absolute difference in 3-year survival was 29.6% less in patients who had weak TTE than in those who had strong TTE on preoperative MRI, with an adjusted hazard ratio of 0.32.

TTE on preoperative MRI was positively correlated with tumour fibrosis and negatively correlated with tumour necrosis on post-hepatectomy specimens, which may be the physiological explanation for this MRI phenomenon. No prior studies have specifically correlated the late gadolinium enhancement in CRCLM with tumour fibrosis, although studies have looked at the correlation between noncontrast MRI signal characteristics of CRCLM and tumour fibrosis [11]. It is well established in the pathology literature that tumour fibrosis in CRCLM is one of the major pathological responses to chemotherapy and the predominant pathological response associated with treatment response and long-term outcomes [4, 5, 12]. Specifically, tumour fibrosis and not tumour necrosis post-chemotherapy is associated with good long-term prognosis [4]. Tumour necrosis is known to be poorly enhancing on contrast-enhanced MRI, which could be a confounding factor for measurement of TTE [13]. In addition, tumour necrosis would be inversely correlated with tumour fibrosis since these variables may demonstrate collinearity.

RECIST is the most commonly used technique for evaluation of chemotherapy response [8]. However, it is a size-based technique that has been shown to correlate poorly with pathological response or long-term survival [14]. Several imaging criteria have been developed to address these limitations, including CT-based morphological criteria, which showed good association with pathological response and survival in the setting of CRCLM treated with bevacizumab-containing chemotherapy [15]. Some authors have assessed the role of imaging techniques in assessing tumour biology, such as DCE-MRI and PET, although these techniques are expensive and time-consuming and are less routinely performed in the clinical setting [15‐18].

Our study had several limitations, mostly related to its retrospective nature. There was variability in the timing of MRI in relationship to the administration of chemotherapy and the type of chemotherapy administered. Radiological-pathological correlation was limited by sampling error, which could decrease Spearman correlation particularly in tumours with significant heterogeneity, and not all lesions could be matched on a per-lesion basis, which could lead to selection bias. Additionally, tumour fibrosis can be seen in CRLM even in patients without chemotherapy [19]. These confounding factors may contribute to the relatively weak correlation between tumour fibrosis and TTE observed in our study.

Although 10-min delayed-phase imaging is part of our institution’s routine clinical liver MRI protocol, the addition of a 10-min delayed phase scan may impede workflow and be a limitation at institutions that only perform imaging to 3- or 5-min post-contrast. We performed TTE analysis at 10-min delayed phase based on the cardiac MRI literature, which has shown that fibrosis is best seen between 10 and 30 min [20, 21]. However, it is unclear whether this is also the case for tumour fibrosis within CRCLM and further studies should be performed to determine whether a 3-5-min delay is also sufficient.

Because MRIs were not obtained for the purpose of measuring CNR, several technical confounders including magnetic field strength, use of phased-array surface coils, and presence of diffuse liver disease may have affected our results [22, 23].

Our study, while important, demonstrates the need for additional prospective studies to confirm the results, for external validation, and to determine its potential clinical impact. Prospective studies are also required to determine the role of specific chemotherapy regimens analysing pre- and post-treatment MRI scans, to confirm correlation between MRI signal and fibrosis using registered, high-resolution, radiological-pathological techniques, to optimise selection of target lesions, and to optimise measurement of TTE through T1 signal mapping. In some patients with multiple CRCLMs, the enhancement pattern of different lesions can be heterogeneous. The presence of heterogeneous target lesions did not affect our results on our post-hoc sensitivity analysis, likely because of the relatively small proportion of patients with heterogeneous target lesions (approximately 12-15% of our patient cohort). However, further studies are required to determine the optimal method of measuring TTE in these patients. Although reliability between raters in our study was very good, development of standardised semi-automated techniques may further improve reliability.

As many patients are now staged using MRI with hepatobiliary-specific contrast agents, investigating the relationship between late-phase enhancement of CRCLM and tumour fibrosis and survival may also be valuable [24].

In conclusion, this article presents the first study to provide evidence that late gadolinium MRI enhancement of tumours post-chemotherapy is associated with tumour fibrosis and overall survival post-hepatectomy in patients with CRCLM. Target tumour enhancement on MRI may be a useful tool for risk-stratification. Further studies are required for external validation and to assess its potential clinical impact.