Abstract
OBJECTIVE: To obtain a summary positive predictive value (sPPV) of contrast-enhanced CT in determining resectability. METHODS: The MEDLINE and EMBASE databases from JAN2005 to DEC2015 were searched and checked for inclusion criteria. Data on study design, patient characteristics, imaging techniques, image evaluation, reference standard, time interval between CT and reference standard, and data on resectability/unresectablity were extracted by two reviewers. We used a fixed-effects or random-effects approach to obtain sPPV for resectability. Several subgroups were defined: 1) bolus-triggering versus fixed-timing; 2) pancreatic and portal phases versus portal phase alone; 3) all criteria (liver metastases/lymphnode involvement/local advanced/vascular invasion) versus only vascular invasion as criteria for unresectability. RESULTS: Twenty-nine articles were included (2171 patients). Most studies were performed in multicentre settings, initiated by the department of radiology and retrospectively performed. The I(2)-value was 68%, indicating heterogeneity of data. The sPPV was 81% (95%CI: 75-86%). False positives were mostly liver, peritoneal, or lymphnode metastases. Bolus-triggering had a slightly higher sPPV compared to fixed-timing, 87% (95%CI: 81-91%) versus 78% (95%CI: 66-86%) (p = 0.077). No differences were observed in other subgroups. CONCLUSIONS: This meta-analysis showed a sPPV of 81% for predicting resectability by CT, meaning that 19% of patients falsely undergo surgical exploration. KEY POINTS: • Predicting resectability of pancreatic cancer by CT is 81% (95%CI: 75-86%). • The percentage of patients falsely undergoing surgical exploration is 19%. • The false positives are liver metastases, peritoneal metastases, or lymph node metastases.