Objective: The purpose of this study was to evaluate the performance of systematic MRI with DWI for the detection of liver metastases (LM) in patients with potentially resectable pancreatic ductal carcinoma and normal liver findings at CT.
Subjects and methods: Patients with potentially resectable pancreatic ductal carcinoma and a normal liver at CT were enrolled in a prospective multicenter study between March 2011 and July 2013 and underwent preoperative MRI. The reference standard was pathologic analysis of detected hepatic lesions.
Results: A total of 118 patients were enrolled. MRI depicted liver lesions that were not visible at CT in 16 patients. All lesions were visualized both with and without DWI. Lesions were LM in 12 (10.2%) patients and were confirmed in seven patients by preoperative biopsy, four by intraoperative frozen section, and one at 6-month follow-up evaluation after pancreatic resection. All but one liver metastatic lesion diagnosed with MRI were smaller than 10 mm. Four of 118 (3.4%) patients had a false-positive diagnosis of LM at MRI and remained LM free after a follow-up period of 24 months or longer. Three of 102 (2.9%) patients with normal MRI findings had subcapsular LM that were diagnosed intraoperatively. At follow-up, 99 of 118 (83.9%) patients were LM free after a mean of 24 months. The patient-based sensitivity of MRI for the detection of LM was 80.0% (95% CI, 51.9-95.7%); specificity, 96.1% (95% CI, 90.4-98.9%); positive predictive value, 75.0% (95% CI, 47.6-92.7%); and negative predictive value, 97.1% (95% CI, 91.6-99.4%).
Conclusion: Compared with CT, preoperative MRI improves the detection of LM in patients with potentially resectable pancreatic ductal carcinoma and may change management and the rate of unnecessary laparotomy and pancreatectomy for 10% of patients.
Keywords: DWI; MRI; liver metastases; pancreatic ductal carcinoma; surgery.