We searched the electrical medical records at out institution to identify the study patients. The eligibility criteria for the main investigation included the following: (1) patients who were histopathologically diagnosed with PDA between January 2005 and December 2012; (2) patients identified to have potentially resectable PDA on CT. One thousand eight hundred fifty-two consecutive patients were newly diagnosed with PDA during this period. All of them underwent CT imaging for work-up of PDA. Our standardized clinical reports on staging CT for PDA, which were made by experienced board-certified abdominal radiologists, were used to identify relevant patients. Two radiologists in consensus reviewed the imaging findings of the baseline CT with reports, and subsequently classified PDA patients into four categories (resectable, borderline resectable, unresectable, and indeterminate state) based on the 2015 NCCN practice guidelines, in order to determine the surgical resectability [12]. The potentially resectable PDA in our study included resectable, borderline resectable, and indeterminate state. Patients were excluded if they had the following: (1) incomplete clinical or radiological data (n=566), (2) unresectable state on CT imaging review (n=946), (3) non-contrast CT or MRI (n=22), and (4) age more than 80 years (n=20). Hence, a total of 298 patients (165 men, 133 women; mean age, 62 years; range, 27 to 80 years) were included in the main study population with potentially resectable PDA on CT. Of these, 82 patients only underwent CT (CT group) and 216 patients underwent MRI within 4 weeks after CT acquisition (CT+MR group). The recommendation regarding the subsequent MRI was made in consensus at multidisciplinary pancreatobiliary meeting, but the final decision was left to the attending physician’s judgment. Clinical information including patient demographics, laboratory data, treatment methods, and histopathology reports were reviewed using electronic medical records. Of the 298 patients, 161 (54%) who received curative surgical resection were included for subgroup analysis to evaluate the recurrence-free survival and time to liver metastasis after surgery. The flowchart of the study population is presented in Fig. 1.

All patients underwent CT examinations with either a 16-channel or 64-channel scanner (Sensation 16 or Sensation 64, Siemens, Erlangen, Germany; Brilliance 64, Philips Healthcare, Cleveland, OH). After obtaining non-contrast CT images, contrast-enhanced pancreatic CT were performed with intravenous administration of nonionic contrast medium (Ultravist 300, Schering, Berlin, Germany). Pancreatic and portal venous phase imaging was obtained by adding 18 seconds to the time of peak abdominal aortic enhancement calculated at the hepatic hilum, and 18 seconds to the end of the pancreatic phase, respectively. In 38 patients, only the portal venous phase CT was obtained. The scanning parameters were as follows: beam collimation, 0.75 and 0.625 mm; slice thickness, 5 and 3 mm; reconstruction interval, 5 and 3 mm; rotation time, 0.5 seconds; effective tube current-time charge, 150-250 mAs; 120 kVp for the 16- and 64-channel scanner, respectively.

MRI was performed by using a 1.5-T (Intera Achieva, Philips Medical Systems, Best, Netherlands) or 3.0-T scanner (Magnetom Trio Tim, Siemens Medical Solutions, Erlangen, Germany) using 4- or 16-channel torso-array coil. All patients underwent baseline non-contrast and post-contrast dynamic MRI. Both single-section and navigator-triggered MRCP imaging were performed. Dynamic T1-weighted imaging was obtained after administering one of the two contrast materials (Primovist, Bayer Schering Pharma, Berlin, Germany; Dotarem, Guerbet, France): a bolus injection of 0.1 mL/kg gadoxetic acid (n=186) at a rate of 1.0 mL/sec or 0.2 mL/kg gadoterate meglumine (n=30) at a rate of 2.0 mL/sec, followed by a 20 mL saline flush using a power injector. The peak time of the time-density curve to determine image acquisition timing was achieved using a test-bolus technique in which 1 mL of gadoxetic acid or of gadoterate meglumine was injected with a saline flush. The arterial phase began 2 or 3 seconds after the peak aortic enhancement was determined. Subsequent portal (50 seconds) and transitional (3 minutes) phase images were obtained. Additional hepatobiliary phase (20 minutes) images were obtained when gadoxetic acid (hepatocyte-specific agent) were used. Diffusion-weighted imaging was obtained at b values of 0, 50, 400, and 800 sec/mm². In our institution, both hepatocyte-specific contrast agent (gadoxetic acid) and diffusion weighted imaging (DWI) were introduced on February 2008. Thereafter, they became the routine protocol. Detailed MR parameters are provided in S1 Table.

Two, board-certified abdominal radiologists (with 6 and 11 years of experience in abdominal imaging, respectively) in consensus retrospectively reviewed the all baseline CT images (n=298 patients) with reference to prospectively interpreted reports at Picture Archiving Communication System. They classified the patients into three categories (resectable, borderline resectable, and unresectable states) based on the 2015 NCCN, as defined in S2 Table. Patients who did not belong to above the three NCCN categories (i.e., patients with hepatic lesions considered too small to characterize or lesions which are difficult to categorize between benign or malignant lesions in any organ) were classified as an indeterminate state. After four weeks, the same radiologists reviewed and classified the MRI (n=216 patients) in the same manner as CT. While reviewing the MRI, they simultaneously referred to the CT imaging findings as in routine clinical practice. The radiologists were informed that all patients had PDA, but they were blinded to all other things, including clinical and laboratory findings and the histo-pathological stage.

The primary outcome of this study was change in surgical resectability after additional MRI in patients with potentially resectable PDA on CT. Secondary outcomes included overall survival rates for all patients, and recurrence-free survival rates and time to liver metastases for patients receiving curative surgical treatment, between the CT only and CT+MR groups. The index date was defined as the date on which the patient underwent baseline CT for the diagnosis of PDA. Patients were followed up from the index date till death or the last follow-up date. Each patient’s overall and recurrence-free survival was calculated from the index date till death, date of last follow-up evaluation, or confirmation of recurrence. To substantiate the completeness of the follow-up data, information about vital status was obtained from the National Population Registry of the Korea National Statistical Office using unique personal identification numbers. Curative-intent treatments were defined as curative surgical resection regardless of neoadjuvant concurrent chemoradiotherapy. The routine follow-up protocol included serum carbohydrate antigen (CA) 19-9 and contrast-enhanced CT. Recurrence of PDA was defined to have occurred when local recurrence or distant metastatic lesions were first evident in the medical record, based on a combination of follow-up CT imaging, CA 19-9 level, histological confirmation, or physical examination. Time to liver metastases was calculated from the date of curative surgery to the initial presentation date of liver metastases on follow-up CT among patients receiving curative surgical treatment. Subgroup analysis was performed for patients in the CT+MR group. They were divided into persistently potentially resectable and unresectable states based on the CT+MR findings, and the overall and recurrent-free survival was compared between the two groups.

Clinical characteristics of all patients were compared between the CT only and CT+MR groups using the Student t test for continuous variables and the chi-square test for categorical variables. Kaplan-Meier survival analysis and Cox proportional hazards regression were used to assess the impact of additional MRI on overall survival and recurrence-free survival. Time to liver metastases for the CT and CT+MR groups were compared by using Mann-Whitney U test.

Statistical analyses were performed by our statistician using the R software ver. 3.3.1. (R Foundation for Statistical Computing, Vienna, Austria). All reported p-values were two sided with a significant level of less than 0.05.

This historical cohort study was approved by the institutional review board of Severance Hospital (IRB No. 2016-1611) that is a tertiary referral center with high volume of pancreatic cancer, and the requirement for informed consent from patients was waived.

Clinical characteristics of the patients are summarized in Table 1. The proportion of patients receiving neoadjuvant concurrent chemoradiotherapy before surgical resection was significantly higher in the CT group (53.1%) than in the CT+MR group (30.4%) (p=0.006). The other baseline characteristics of patients in the CT and CT+MR groups were not significantly different. The proportion of patients undergoing additional MRI was 72.4% (142 of 196 patients) for the resectable state, 66.2% (49 of 74 patients) for the borderline resectable state, and 89.3% (25 of 28 patients) for the indeterminate state at CT classification, without statistical difference (p=0.067).

Of the 216 patients in the CT+MR group, a total of 31 patients (14.4%) who had potentially resectable states on CT were changed to have unresectable state on additional MRI (Table 2). Liver metastases (n=16), major vascular invasion (n=13), enlarged para-aortic lymph node (n=1), and peritoneal seeding (n=1) were additionally found by MRI. Of the 16 patients with liver metastases diagnosed on MRI, nine patients had negative-liver-on-CT (4.7%, 9 of 191 resectable or borderline resectable group on CT) (Fig. 2) and seven patients had diminutive indeterminate hepatic lesions on CT (35%, 7 of 20). Of the 13 patients with unresectable state due to major vascular invasion on MRI, three patients had no major vessel invasion on CT (1.4%, 2 of 142 resectable group and 4%, 1 of 25 indeterminate group on CT) and 10 patients had borderline vascular invasion on CT (20.4%, 10 of 49 borderline group on CT). Two of five patients with indeterminate extrahepatic lesions (omental haziness and preaortic lymph node) other than indeterminate hepatic lesions on CT were changed to surgically resectable state after MRI. After analysis of MRI, 185 patients in CT+MR group were remained as persistently potentially resectable. There was no significant difference in the overall proportion of the patients who received curative surgical treatment between the CT only (49 of 82, 59.8%) and the persistently potentially resectable CT+MR groups (111/185, 60%) (p=0.970). Thirty of 31 patients reclassified into the unresectable state after MRI did not receive curative surgical treatment (96.8%).

The median follow-up period was 1.6 years for the CT group and 1.4 years for the CT+MR group without significant difference between the two groups (p=0.193). The 5-year overall survival rates of the CT only and CT+MR groups were 24.1% and 17.2%, respectively, without statistical difference (p=0.162) (S3A Fig.).

Within the CT+MR group, the 5-year overall survival rates between the potentially resectable (19.2%) and unresectable (4.8%) subgroups were significantly different (p < 0.001) (hazard ratio [HR], 2.27; 95% confidence interval [CI], 1.53 to 3.37; p < 0.001) (Fig. 3). The curative surgical treatment performed in PDA patients was significantly associated with a lower risk of overall patient mortality (HR, 0.39; 95% CI, 0.29 to 0.52; p < 0.001) (Table 3).

A total of 161 patients (54%) received curative surgical resection with 112 patients in the CT+MR group and 49 patients in the CT group. Five-year recurrence-free survival rates were 29.8% and 29.7%, respectively, without statistical difference (p=0.721) (S3B Fig.). However, the median time to liver metastases developed after curative surgery was 4.2 months in the CT group and 9.9 months in the CT+MR group with significant difference between the two groups (p=0.011), although the number of liver metastases confronted during their whole follow-up period was not very different (44 of 112 patients [39.3%] in the CT+MR group and 20 of 49 patients [40.8%] in the CT group) (Fig. 4).