An Independent Study of a Gene Expression Classifier (Afirma) in the Evaluation of Cytologically Indeterminate Thyroid Nodules

Context: Molecular markers hold the promise of improved diagnostic yield in thyroid fine-needle biopsy. The Afirma gene expression classifier (GEC), available commercially, reports a negative predictive value of 94% in the diagnosis of benign nodules after indeterminate cytology. However, there are currently no independent studies of the performance of this assay. Objective: The aim was to assess the performance of the Afirma GEC in an academic medical center. Design: Samples for the GEC were collected according to the manufacturer's recommended protocol from patients undergoing thyroid fine-needle aspiration. We requested GEC analysis on nodules reported cytologically as follicular neoplasm or atypia or follicular lesion of undetermined significance from patients willing to defer surgery. Patients: All patients undergoing thyroid fine-needle aspiration during the study period, whose cytology was reported as follicular neoplasm or atypia of undetermined significance/follicular lesion of undetermined significance, were offered access to the test and recruited to this study. Intervention: Patients whose GEC was benign were offered ultrasound follow-up in lieu of surgery. Those with a suspicious GEC were advised to undergo diagnostic lobectomy. Setting: The study was conducted at a large academic medical center. Main Outcome Measure: We measured the rate of benign and suspicious calls from the Afirma GEC and histological diagnosis after surgery. Results: A total of 72 nucleic acid samples were sent for GEC analysis. In 12 (17%) of these samples, there was insufficient mRNA, leaving 60 Afirma results for analysis. Of these, 16 (27%) were benign, whereas 44 (73%) were suspicious. The rate of confirmed malignancy in GEC-suspicious nodules was only 17%. Conclusion: The Afirma GEC demonstrates a lower than expected rate of benign reports in follicular or Hurthle cell neoplasm and a lower than anticipated malignancy rate within GEC-suspicious nodules. These data suggest that the positive predictive value of the GEC is lower than previously reported and call into question the performance of the test when applied in the context of specialized academic cytopathology.