Clinical Validation of the Afirma Genomic Sequencing Parathyroid Classifier


Interview with Richard T. Kloos, M.D., Veracyte

Dr. Richard Kloos is senior director, endocrinology at Veracyte, where he draws upon his extensive academic and clinical experience in endocrinology to educate physicians about the clinical use and impact of the company’s genomic technology. Here he describes data being presented at the ATA Annual Meeting, which validate the performance of the Afirma Genomic Sequencing Classifier’s (GSC) parathyroid classifier in thyroid cancer diagnosis.

Q: Why is it important to distinguish parathyroid from thyroid samples when evaluating patients with suspected thyroid nodules? 

Dr. Kloos: The parathyroid glands are located next to the thyroid and are sometimes embedded within it, so enlarged and embedded parathyroid glands can be mistaken for thyroid nodules. Parathyroid abnormalities, however, are treated very differently from thyroid nodules and have their own management guidelines. Some parathyroid neoplasms can be left alone, while others require surgery and/or other treatment.

Because parathyroid cells often look like atypical thyroid cells under the microscope, cytopathologists often designate them as indeterminate – typically, as Bethesda III or IV. This can lead to unnecessary thyroid surgery for these patients to determine if cancer is present – and to the morbidity, risks and costs that can accompany such surgery.

Q: How does the Afirma GSC’s parathyroid classifier work?

Dr. Kloos: The Afirma GSC combines RNA sequencing and ensemble machine learning algorithms to assess thyroid nodule fine needle aspiration, or FNA, biopsies and identify benign nodules among those that are indeterminate by cytopathology. The Afirma GSC also includes algorithms that detect the molecular signatures of specific neoplasms – such as unsuspected parathyroid cells – which can change how the patient is managed and treated. The parathyroid classifier was trained with 476 FNA samples and contains 109 genes.

Q: How did you evaluate the performance of the Afirma GSC parathyroid classifier? 

Dr. Kloos: We blindly tested the final classifier on an independent test set of 195 FNA samples that were categorized as Bethesda III (118) or Bethesda IV (77). All of the positive samples had clinical and/or surgical confirmation of the parathyroid etiology, while all negative samples were negative on surgical pathology.

Q: So what did you find? 

Dr. Kloos: Among the 195 test-set FNAs, there were 4 parathyroid samples and the classifier identified all of them correctly, meaning it had 100 percent sensitivity. It also correctly called all 191 non-parathyroid samples as negative, meaning it had 100 percent specificity.


Q: What’s the significance of these findings? 

Dr. Kloos: Our results showed that the Afirma GSC’s parathyroid classifier can help to correctly identify patients who have parathyroid abnormalities, rather than thyroid nodules. This can help these patients potentially avoid an unnecessary surgery, while also getting more appropriate treatment. This provides optimal care for the patient and reduces waste in the healthcare system.

Read the Abstract Here