Scoring system for intraoperative diagnosis of intracranial schwannoma by squash cytology

Objective To assess the utility of a newly developed squash cytology (SC)-based scoring system for accurate intraoperative diagnosis of schwannoma. Methods We first compared SC-based and frozen section (FS) diagnoses with final pathological diagnoses of schwannoma (16 cases), meningioma (39 cases) and low-grade astrocytoma (16 cases). Then, by logistic regression modeling, we identified features of SC preparations that were independently predictive of schwannoma. To develop a diagnostic scoring system, we assigned one point to each feature, and performed receiver operating characteristic analysis to determine the score cut-off value that was most discriminatory for differentiating schwannoma from the other tumour types. We then compared accuracy, sensitivity, and specificity of diagnosis before and after the application of the scoring system. Results Overall diagnostic concordance rates for SC and FS were almost the same, at 73.2% (52/71) and 77.5% (55/71 cases), respectively. Of the 16 SC features entered into the analysis, the following nine were found to independently predict schwannoma, and were thus incorporated into the scoring system: smooth cluster margins, few or no isolated tumour cells, fibrillary stroma, spindle-shaped nuclei, parallel arrangement of stroma, parallel arrangement of nuclei, presence of anisonucleosis, absence of nucleoli, and hemosiderin deposition. A cut-off score of four items yielded the best sensitivity, specificity and predictive values for prediction of schwannoma. Use of the scoring system improved accuracy of intraoperative diagnosis from 80.3% to 94.4%, sensitivity from 56.2% to 93.8%, and specificity from 87.3% to 94.5%. Conclusion Our proposed SC-based scoring system will increase accuracy of intraoperative diagnosis of schwannoma vs non-schwannoma tumours.