Back to Annual Meeting Posters

Background: Utility of positron emission tomography (PET) as an adjunctive imaging modality to CT or MRI in evaluating resectability of pancreatic cancer is a subject of controversy. In this study, we seek to assess the utility of PET in identifying occult metastatic disease, as well as evaluate predictive value of maximum standard uptake value (SUV) with respect to tumor resectability and patient survival.
Methods: Cross sectional imaging, clinical course, operative outcomes, and overall survival of all patients who presented with pancreatic adenocarcinoma and had PET scan in workup were reviewed retrospectively. Resectability was assessed based on established criteria. Continuous variables were compared using Student's t-test or ANOVA. Categorical variables were compared using chi-square or Fisher's exact test. Prediction models were constructed using linear or logistic regression where appropriate.
Results: Complete imaging and follow-up data was available for 123 patients evaluated from 2005 to 2011. Of this cohort, 36 patients (29%) were thought to be free of extrapancreatic disease and offered resection, 21 (17%) had metastatic disease, and 66 (53%) were deemed locally advanced and referred for neoadjuvant therapy. PET and CT/MRI were concordant in 108 (88%) cases, however metastatic lesions were identified in 7 (5.6%) patients deemed resectable by CT or MRI. Among those offered immediate resection, 5 (14%) patients had occult metastatic disease identified at diagnostic laparoscopy, including 3 previously identified by nonconcordant PETs and 2 missed by false negative PETs. False positive PETs led to unnecessary procedures delaying surgery for 3 (8.3%) patients who went on to resection. In a cohort of patients thought to be free of metastatic disease, in terms of detecting metastases, overall sensitivity and specificity of PET were 89.3% and 85.1%, respectively, compared with 62.5% and 93.5% for CT and 61.5% and 100.0% for MRI. Positive predictive value and negative predictive value of PET were 64.1% and 96.4% respectively, compared with 75.0% and 88.9% for CT and 100.0% and 91.9% for MRI. Average difference in maximum SUV of resectable and unresectable lesions was not statistically significant (5.65 vs. 6.5, p=0.224), nor was maximum SUV a statistically significant predictor of survival (p=0.18).
Conclusion: PET is a more sensitive modality for identifying metastatic disease than CT or MRI, however, it has a lower specificity and lower positive predictive value. While PET identified an additional 5.6% of patients with occult metastatic disease, it is likely that unresectability would have been established at diagnostic laparoscopy, thus not saving an unnecessary resection. We therefore conclude that PET has limited utility in workup of patients who already undergo CT or MRI as part of initial staging of pancreatic adenocarcinoma.