Molecular-derived estimation of risk of malignancy for indeterminate thyroid cytology diagnoses.

INTRODUCTION

One of the key features of the Bethesda System for Reporting Thyroid Cytopathology is the risk of malignancy (ROM), which guides management for each diagnostic category. However, calculation of the ROM can be challenging for indeterminate diagnoses because only a portion of cases will be resected for cytologic-histologic correlation (CHC) analyses. In the present study, we used the probability of cancer information from ThyroSeq, version 3, reports to calculate the molecular-derived (MD) ROM for indeterminate categories.

MATERIALS AND METHODS

Cytology cases with indeterminate BSRTC diagnoses and adequate molecular test results were retrieved from our cytopathology laboratory for a 12-month period. The probability of cancer information from the ThyroSeq, version 3, molecular reports were tabulated, and the mean ROM was calculated for each diagnostic category. The MDROM included noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) as a "malignant" outcome because it is considered a surgical disease.

RESULTS

A total of 361 cases had adequate material for molecular testing. The diagnostic distribution was as follows: atypia of undetermined significance/follicular lesion of undetermined significance, 271 cases (75.1%), follicular neoplasm/suspicious for a follicular neoplasm, 59 cases (16.3%), and Hürthle cell type/suspicious for a follicular neoplasm, Hürthle cell type, 31 cases (8.6%). The corresponding estimated MDROMs were 14.9%, 32.6%, and 34.4%. A comparison with the CHC data was performed, and the 95% confidence intervals of the MDROMs overlapped well with the 2 endpoint CHC values.

CONCLUSIONS

Calculation of the MDROMs provides a new method to approximate the ROMs of indeterminate diagnoses and has the advantage of potentially evaluating all cases, not just those resected. Furthermore, for those using the same platform, interinstitutional comparisons will be possible.