Background: Clinical management of thyroid neoplasms is based on light microscopic diagnosis, but its accuracy and precision are poorly defined.
Objective: To assess inter- and intraobserver variability of preoperative cytopathologic and postoperative histopathologic thyroid diagnoses.
Design: Samples were collected in a prospective, multicenter trial validating a gene expression classifier between June 2009 and December 2010.
Setting: 14 academic and 35 community clinical sites.
Patients: 653 patients with 776 surgically resected thyroid nodules of 1 cm or greater.
Measurements: Intraobserver concordance among 2 or more central histopathologists who independently read histopathology slides was calculated. Interobserver concordance between the diagnoses made by the central histopathologists and those made by local pathologists were calculated. Intra- and interobserver concordance for cytopathology was similarly calculated by comparing diagnoses made by local pathologists with those made by a central panel of 3 cytopathologists.
Results: Concordance on the histopathologic distinction between benign and malignant diagnoses was 91% comparing local with central histopathologists and 90% comparing 2 central histopathologists. Using the 6-category Bethesda System, 64.0% of diagnoses made by local and central cytopathologists and 74.7% of intraobserver diagnoses were concordant. Central cytopathologists made fewer indeterminate diagnoses than local pathologists (41.2% vs. 55.0%).
Limitations: Many local pathologists did not use the Bethesda System, so their reports were translated to allow comparison. The study required histopathology, and the study population and specimens did not encompass all newly evaluated patients with a thyroid nodule.
Conclusion: Substantial inter- and intraobserver variability exists in the cytopathologic and histopathologic evaluation of thyroid nodules, confirming an inherent limitation of visual microscopic diagnosis.
Primary funding source: Veracyte.