EUS-guided FNA for diagnosis of solid pancreatic neoplasms: a meta-analysis

Background: Preoperative diagnosis of solid pancreatic lesions remains challenging despite advancement in imaging technologies. EUS has the benefit of being a minimally invasive, well-tolerated procedure, although results are operator-dependent. The addition of FNA (EUS-guided FNA) provides samples for cytopathologic analysis, a major advantage over other imaging techniques.

Objective: To determine the diagnostic accuracy of EUS-FNA for pancreatic cancer.

Design: This is a meta-analysis of published studies assessing the diagnostic capability of EUS-FNA. Relevant studies were identified via MEDLINE and were included if they used a reference standard of definitive surgical histology or clinical follow-up of at least 6 months.

Main outcome measurements: Data from selected studies were analyzed by using test accuracy meta-analysis software, providing a pooled value for sensitivity, specificity, diagnostic odds ratio, and summary receiver operating characteristic curve. Cytology results were classified as inadequate, benign, atypical, suspicious, or malignant. Predefined subgroup analysis was performed.

Results: Thirty-three studies published between 1997 and 2009 were included, with a total number of 4984 patients. The pooled sensitivity for malignant cytology was 85% (95% confidence interval [CI], 84-86), and pooled specificity was 98% (95% CI, 0.97-0.99). If atypical and suspicious cytology results were included to determine true neoplasms, the sensitivity increased to 91% (95% CI, 90-92); however, the specificity was reduced to 94% (95% CI, 93-96). The diagnostic accuracy of EUS-FNA was enhanced in prospective, multicenter studies.

Limitation: Publication bias was not a significant determinant of pooled accuracy.

Conclusion: This meta-analysis demonstrates that EUS-FNA is a highly accurate diagnostic test for solid neoplasms of the pancreas and should be considered when algorithms for investigating solid pancreatic lesions are being planned.