Predictors of malignancy in patients with suspicious or indeterminate cytology on pancreatic endoscopic ultrasound-guided fine-needle aspiration: a multivariate model.

OBJECTIVE

The aim of this study was to evaluate predictors of malignancy in pancreatic lesions with suspicious or indeterminate endoscopic ultrasound fine-needle aspiration (EUS-FNA) cytology.

METHODS

Suspicious/indeterminate EUS-FNA cytology was identified from our database. Stable imaging, benign pathology, or survival for 12 months after EUS-FNA was considered benign. Diagnosis of malignancy was based on positive pathology, local invasion/metastasis on imaging, or death within 12 months from cancer-associated causes. Univariate analysis was performed to compare variables between benign and malignant lesions. Multivariate analysis (covariates: age [<70 or ≥70], appearance [solid/cystic], size [<20 or ≥20 mm], and serum CA19-9 [<40 or ≥40]) was performed using binary logistic regression.

RESULTS

There were 81 patients with suspicious or indeterminate cytology. Final diagnosis was cancer in 55 (67.9%) of 81 (adenocarcinoma in 40/81 [49.4%], "other neoplasms" in 15/81 [18.5%]) and benign in 26 (32.1%) of 81. Univariate analysis revealed a difference in age, lesion size, solid/cystic characteristics, and serum CA 19-9 between benign and malignant lesions. Only elevated serum CA 19-9 was independently associated with a diagnosis of malignancy on multivariate analysis.

CONCLUSIONS

Age, lesion size, and solid/cystic characteristics on EUS were not independently associated with cancer. Pancreatic lesions with suspicious/indeterminate cytology and elevated serum CA 19-9 should be considered as concerning for a malignant diagnosis.