A total of 104 fresh specimens from 28 patients who underwent TSS for PitNETs between April and June 2022 at Moriyama Memorial Hospital were examined for intraoperative pathological consultation. Our clinical team always uses IPD to decide regarding clinically effective resection. All cases had one specimen obtained from the clinically central part of the tumor (clT) to identify the histological variation of the tumors and subsequently, other specimens from clinically surgical margins (clSM), which were defined as pseudo-membranous tissue composed of PitNET and compressed anterior lobe tissue, or from sites where the surgeons could not certify whether tumor components existed, sometimes with irregular fibrosis.

Coated glass slides (MAS-01; Matsunami Glass Industry, Osaka, Japan) were used for both the TIC and the FS procedure. At first, small and fresh tissue samples retrieved during TSS were touched onto dry microscopic glass slides for imprint preparation, and the slides were immediately fixed with 95% alcohol. During the imprinting procedure, at least two surfaces were attached and rolled to examine circumferentially (Fig. 1). After cytology processing, the specimens were mounted in a compound (Tissue Tek O.C.T Compound; Sakura Finetek, Tokyo, Japan), rapidly frozen in isopentane chilled by dry ice, and sectioned at 5 µm. The specimens from both the TIC and FS procedures were stained using the rapid hematoxylin and eosin procedure, dehydrated, permeabilized, and sealed with a glass cover. After the FS procedure, the specimens, excluding the pure tumors for deep freeze storage, were re-fixed with formalin and processed into paraffin-embedded sections for re-diagnosis.

In our touch imprint cytology method, we rolled the specimens and touched two or more surfaces with the microscope glass slides

All slides were independently examined and discussed by a pathological trainee (NT) and a specialized pathologist (NI) with experience in diagnosing over 3,000 PitNET cases.

The cytological findings were classified into the following three categories according to the cell imprint pattern (Fig. 2): pattern cyA, rich and diffusely spread neuroendocrine cells (Fig. 2b); pattern cyB, a small number of scattered endocrine cells (Fig. 2d); and pattern C, no apparent endocrine cells (Fig. 2f). Patterns cyA and cyB were further classified into three subcategories based on the morphology of the endocrine cells: 1; PitNETs, 2; adenohypophyseal cells, and 3; difficult to determine (Fig. 3).

a–f Microscopic images of hematoxylin and eosin staining of touch imprint cytology preparation and frozen section. All specimens were obtained with × 10 magnification. Both a and b were from the same specimen from hT. c and d were from hB. e and f were from hN. a, c, and e were from frozen section, and b, d, and f were touch imprint cytology and classified as follows; b as cyA1, d as cyB1, and f as cyC

A breakdown of cytological classification of specimens

Histologically, with FS on the site of IPD, each specimen was divided into one of the three groups: “histologically tumor (hT)” (tumor component only), “histologically boundary (hB)” (both PitNET and a normal component), and “histologically normal (hN)” (no tumor component). Subsequently, they were evaluated histologically using the formalin re-fixed specimens after FS. The diagnostic accuracy of TIC was validated based on histological findings using both frozen and formalin re-fixed specimens for the final diagnosis.

Based on these definitions, the positivity, sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of TIC were evaluated. The Kappa coefficient was used for diagnostic concordance between TIC and FS histology. Statistical analyses were performed using EZR ver. 1.54 [9]. The Kappa coefficient was assessed per Koch et al.’s definition [10]. Although cyA1 and cyB1 were cytological PitNET-positive specimens, the possibility that a small number of PitNET cells in the surgical field were attached to the surface of the specimen could not be ruled out in cyB1. Therefore, we raised two statistical questions that defined only cyA1 as cytologically positive and both cyA1 and cyB1 as cytologically positive.