ACCURACY OF KI-67 MEASUREMENT ON EUS-FNA CYTOLOGICAL SAMPLES VS SURGICAL SPECIMENS: A SINGLE CENTRE EXPERIENCE

Aims:

The number of pre-operative EUS-FNA for suspected pancreatic neuroendocrine tumors (pNETs) is increasing. EUS-FNA provides pre-operative cytological diagnosis and information about the biological behaviour of the neoplasia. Measurement of Ki-67 expression on EUS-FNA specimen is feasible, but it is still undefined how this measure could reflect the value of Ki-67 expression assessed on surgical specimens. We report our single center experience, describing the feasibility of measurement of Ki-67 expression on cytological samples and evaluating its accuracy in comparison to Ki-67 expression on surgical specimens, with the purpose of demonstrating the reliability of EUS-FNA for guiding, alone, the clinical management of patients unfit for surgery.

Methods:

We made a retrospective revision of all cases of pNETs diagnosed through EUS-FNA that underwent surgery between August 2003 and October 2017. Ki-67 on cytological samples was compared with the histological Ki-67 expression on surgical specimens evaluating agreement rates and the possible impact on clinical management using different cut-off values.

Results:

48 patients (22 M, 26 F) with an EUS-FNA diagnosis of pNET underwent surgical resection. Comparison between cytological and histological Ki-67 measurement was possible in 44/48 cases. Using the ENETS grading (G1< 2% and G3 > 20%), overall agreement rate was 32/44 (72,7%), while agreement rates with 2% only (G1 vs. G2-G3) and with 20% only (G1-G2 vs. G3) were 31/44 (70,4%) and 43/44 (97,7%) respectively. A new grading with G1-G2 cut-off set at 5% instead of 2%: with this classification, agreement rates were substantially comparable: 33/44 (overall), 33/44 (G1 vs. non-G1), 43/44 (G3 vs. non-G3).

Conclusions:

Measurement of Ki-67 expression on cytological samples is easily realized on high-cellularity specimens. Although its sub-optimal accuracy considering low cut-off values (2 or 5%), agreement with definitive histological measurement is very high (97,7%) considering high cut-off value (20%), probably the most important cut-off able to influence clinical decision in pNETs management.