Added value of 3T MRI and the MRI-halo sign in assessing resectability of locally advanced pancreatic cancer following induction chemotherapy (IMAGE-MRI): prospective pilot study

Approximately 30–40% of the patients with pancreatic ductal adenocarcinoma (PDAC) present with locally advanced pancreatic cancer (LAPC) [1]. LAPC is characterized by extensive vascular tumor involvement prohibiting an upfront curative-intent resection [2]. The introduction of FOLFIRINOX (i.e., a combination of 5-fluorouracil, leucovorin, irinotecan, and oxaliplatin) and gemcitabine-nab-paclitaxel induction chemotherapy has resulted in the opportunity to achieve local and systemic control of LAPC. As a consequence, a resection is possible in 16–26% of patients with LAPC [3‐5], associated with survival rates comparable to those obtained in (borderline) resectable disease [6‐9].

Patient selection for surgical exploration is based on both tumor biology [10] and anatomical staging [11, 12]. Unfortunately, contrast-enhanced computed tomography (CE-CT) imaging is inaccurate to determine the resectability of LAPC following induction therapy because it cannot differentiate between vital tumor tissue and fibrosis [13], whereby the sensitivity and specificity on the resectability in patients with PDAC who underwent resection after preoperative chemo(radio)therapy are 45–78% and 60–85%, respectively [14, 15]. Consequently, local response to chemotherapy may be underestimated, which might result in missed opportunities for resection or, conversely, futile exploratory laparotomy [16]. Improvement of preoperative imaging modalities is urgently needed to reduce futile surgery and prevent the associated poor short- and long-term outcomes [17].

In pancreatic cancer, contrast-enhanced magnetic resonance imaging (CE-MRI) is generally recommended only as additional imaging next to CE-CT in case of a poorly visible tumor and to exclude liver metastases [18, 19]. Diffusion-weighted magnetic resonance imaging (DWI-MRI) has shown promising results in determining tumor response after chemotherapy in PDAC, focusing on the apparent diffusion coefficient (ADC) and its possible correlation with histological tumor response [20‐22]. However, the role of CE-MRI and DWI-MRI for morphological response evaluation after chemotherapy is poorly investigated to our knowledge [16, 23, 24], especially when focusing on the perivascular zone of the tumor to depict a “halo sign,” which is considered to reflect tumor regression [25, 26]. A halo sign has been studied and described with CE-CT [25, 26], but not with MRI. A halo sign could be a valuable marker for remaining vascular tumor involvement after induction chemotherapy and hereby could improve the preoperative anatomical staging and subsequent patient selection for surgery [19, 25‐27]. As other studies have shown that DWI-MRI can differentiate vital tumor tissue from fibrosis in rectal cancer after chemoradiation therapy and CE-MRI is assumed to differentiate better between different tissue types [28], we hypothesized this may also be the case in patients with PDAC.

Therefore, this prospective pilot study explored the value of CE-MRI and DWI-MRI separately, compared with standard CE-CT to determine tumor size and vascular involvement of LAPC following induction chemotherapy.

This prospective observational pilot study was performed following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement guideline [29].

The Medical Ethics Committee from the Amsterdam UMC (location Academic Medical Center) approved this prospective study (registration number: NL5665.018.16). Before inclusion, all patients provided written informed consent.

Primary endpoints included tumor size and vascular tumor involvement after induction chemotherapy on CE-CT versus CE-MRI and CE-CT versus DWI-MRI, including the presence of a halo sign. The secondary endpoints included subjective image quality of CE-CT versus CE-MRI and CE-CT versus DWI-MRI.

Out of 22 eligible patients, two patients were not able to undergo an MRI because of claustrophobia, leading to a final cohort of 20 patients. Except one patient, all MRI scans were performed after a CE-CT scan. The mean time interval between CE-CT and MRI was 25 days (± 14).

The findings of this first prospective pilot study, investigating the added value of CE-MRI in assessing tumor response in LAPC in addition to CE-CT, suggest less tumor involvement and smaller tumor size after induction chemotherapy on CE-MRI by the more frequently seen halo sign on CE-MRI (53%) compared to CE-CT (11%). Patients with a CE-MRI-halo sign appeared to have a non-significant higher rate of R0 resection in comparison to patients without an CE-MRI-halo sign, but this should be seen as purely hypothesis generating and should be confirmed in larger studies. In contrast, vascular involvement was mainly not assessable on DWI-MRI and thus may not have clinical utility for local, morphological staging purposes.

MRI is mentioned as a potential functional imaging modality (e.g., DWI-MRI) to predict oncological outcomes (e.g., histopathological response, R0 resection, disease-free, and overall survival) after chemo(radio)therapy in patients with pancreatic cancer [16, 23, 39]. On the other hand, the role of MRI as morphological imaging modality in these patients is unclear. A recent review paper from the Society of Abdominal Radiology stated that the limitations of morphological imaging (both CT and MRI scan) after chemo(radio)therapy for PDAC are well established [23]. Of note, this view on MRI seems to be based on a meta-analysis [40] from the era where preoperative chemo(radio)therapy was not widely implemented yet. To our best knowledge, in-depth literature about the value of MRI for morphological restaging after chemo(radio)therapy for PDAC is still lacking.

The present study suggests that CE-MRI might have added value when restaging LAPC following induction chemotherapy, next to CE-CT. The presence or absence of a halo sign on CE-MRI could discriminate between vital tumor tissue and fibrosis after induction therapy although further larger prospective studies are needed to prove this suggestion. This hypothesis is carefully strengthened by differences in vascular involvement between CE-CT scan and CE-MRI, where vascular involvement seemed generally less on MRI in the presence of an MRI-halo sign, whereas the opposite appeared to be true in the absence of an MRI-halo sign (i.e., increased vascular involvement compared with CT scan). The finding that the halo sign was more frequently seen on CE-MRI than on CE-CT may suggest that MRI might be more sensitive to assess post-chemotherapy resectability, but this needs to be proven in larger studies.

This study confirms that DWI-MRI is insufficient for morphological staging since tumor delineation and vascular tumor involvement were hardly assessable. This may not come as a surprise, given the generally low spatial resolution of this modality. Interestingly, also the borders of the tumors were difficult to distinguish from the normal pancreas parenchyma since the pancreas often demonstrated diffusion restriction. This phenomenon was also observed in one of our previous series, investigating different models like DWI-MRI models in patients with (borderline) resectable pancreatic cancer who were treated with neoadjuvant gemcitabine plus hypo-fractionated radiation [21]. Possibly, this is a result of chemotherapy [41], but this cannot be confirmed with the present study because no MRI scan was made before chemotherapy.

Whereas the current results suggest the potential value of CE-MRI for morphological response evaluation after induction chemotherapy, the benefits of CT scan over MRI comprise a higher contrast resolution (that is key for adequate tumor delineation), shorter scan time (so fewer motion artifacts and less burden for patients), and lower costs [18, 23]. Taking this into account, optimization of CE-CT interpretation (especially regarding arterial involvement) could be helpful [19], next to investigating the additional or alternative role of other imaging modalities such as CE-MRI. In recent years, various modified criteria and composite scores are proposed, aiming to optimize the predictive value of CT scan after preoperative therapy. Noda and colleagues designed a composite score based on the presence and characteristics of arterial involvement and resectability status before and after preoperative therapy [42]. This score was correlated with both R0 and overall survival. However, this score was biased since it was developed in a cohort that underwent surgical resection [42]. In addition, Yang et al. investigated the diagnostic accuracy of different CT criteria (National Comprehensive Cancer Network [NCCN] versus modified criteria) based on published literature [15]. Modified resectability criteria involved more nuances in the assessment of vascular involvement, such as changes in perivascular tissue compared to the pretreatment situation. Meta-analysis demonstrated a higher diagnostic accuracy to predict R0 resection when using the modified criteria (0.78 [95% CI 0.74–0.82] versus 0.67 [95% CI 0.63–0.71]) [15]. Ahmed and colleagues showed that more detailed criteria involving the degree and length of involvement and vessel deformation are reproducible and sufficient predictors for R0 resection [31, 43]. In addition to such modified criteria, the change of tumor attenuation on CE-CT scan may be useful to predict R0 resection [44] as well as the tumor homogeneity that seems to correlate with disease-free and overall survival [45]. Further nuance in resectability from advanced PDAC with major arterial tumor involvement based on CT imaging can be made with the presence of a string versus halo sign, possibly differentiating between the presence and absence of vessel wall invasion [26, 27]. In addition, the halo sign on CE-CT as sign of absent vascular invasion has also been described for venous involvement (25). In the future, the extraction of radiomic features could further improve the judgment of resectability after chemo(radio)therapy [46, 47].

The often-overestimated vascular tumor involvement on CE-CT scans after chemotherapy might, to a yet undefined extent, be more accurately assessed with (additional) CE-MRI. If this will be confirmed by more robust studies, CE-MRI could possibly improve the anatomical restaging of LAPC after induction chemotherapy, allowing for better selection for oncological and surgical treatment. Further prospective observational studies are required to investigate the diagnostic accuracy and clinical value of CE-MRI in comparison/next to CE-CT scan to predict R0 resection and to investigate the meaning of a halo sign by involving the histopathological ground truth. Interestingly, an association of CE-MRI-halo sign and tumor marker patterns may be valuable in the development of prognostic models. In addition, other predictors for resectability on CE-MRI have to be investigated (e.g., tumor shrinkage and [changes in] contrast enhancement) as well as the predictive value of DWI-MRI with ADC quantification. Until the role of CE-MRI is validated, the above-mentioned CT criteria might support the anatomical restaging with CE-CT scan. Additionally, patient selection could be further improved by implementing the use of PET combined with CT or MRI before and after induction chemo(radio)therapy for local response evaluation [8, 39, 48] and intraoperative ultrasonography during an exploratory laparotomy to assess the (extent of) vascular tumor involvement [49, 50].

The present study should be interpreted in the light of several limitations and strengths. First, no MRI was performed before induction chemotherapy, so the findings cannot be related to the baseline situation. Second, the sample size is small since this was a pilot study, giving limited value to the calculated diagnostic performance parameters. Since only 2 patients had a CE-CT-halo sign, calculating the diagnostic performance and a subsequent comparison with the diagnostic performance of the CE-MRI-halo sign was considered not reliable. In addition, the diagnostic performance of the DWI-MRI was not calculated, because of the evident lack of value for morphological restaging. Third, the interobserver variability was not assessed since the imaging was immediately scored in consensus. However, as the CE-MRI-halo sign is a new phenomenon, our radiologists chose to work in consensus in order to learn how to interpret the sign and reducing the impact of confounders as a consequence of anatomy-based categorization of vascular involvement. To the best of our knowledge, this is the first prospective study that investigated the role of MRI for local morphological restaging in patients with LAPC patients after induction chemotherapy.