Conference Coverage

New and improved classifiers may sharpen thyroid nodule diagnosis


Key clinical point: New iterations of molecular classifiers show promise for better preoperative diagnosis of thyroid nodules.

Major finding: ThyroSeq v3 had a negative predictive value of 97%. Specificity was an absolute 47% greater for Afirma GSC with Hurthle-specific classifiers than for Afirma GEC. The Afirma GSC with a medullary thyroid cancer classifier had 100% sensitivity and specificity.

Data source: Validation studies of the ThyroSeq v3 classifier (257 samples), the Afirma GSC with Hurthle-specific classifiers (186 samples), and the Afirma GSC with a medullary thyroid cancer classifier (211 samples).

Disclosures: Dr. Nikiforov disclosed that he is owner of an IP for ThyroSeq, and that his laboratory has a contract to offer the test commercially. Dr. Duh disclosed that he had no relevant conflicts of interest. Dr. Randolph disclosed that he had no relevant conflicts of interest.


AT ATA 2017

– Several new and improved molecular classifiers show good performance for preoperatively assessing the nature of thyroid nodules, including histologic subsets that continue to pose diagnostic challenges, according to a trio of studies reported at the annual meeting of the American Thyroid Association.

ThyroSeq v3 classifier

In a prospective, blinded, multi-institutional study, investigators validated the ThyroSeq v3 genomic classifier, which uses next-generation sequencing to test for mutations, fusions, gene expression alterations, and copy number variations in 112 genes.

The validation cohort consisted of 234 patients from 10 centers who had thyroid nodules with Bethesda III to V cytology and known surgical outcome, with central pathology review, and successful molecular testing. In total, they had 257 fine needle aspiration samples.

Of the 247 samples from nodules having Bethesda III or IV cytology – those of greatest interest – 28% were cancer or noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP), reported senior author Yuri Nikiforov, MD, PhD, professor of pathology and director of the division of molecular & genomic pathology at the University of Pittsburgh Medical Center. “Both cancer and NIFTP are surgical diseases, so we felt they belong in one group,” he noted.

Dr. Yuri Nikiforov, professor of pathology and director of the division of molecular & genomic pathology at the University of Pittsburgh Medical Center

Dr. Yuri Nikiforov

Among the Bethesda III or IV samples, ThyroSeq v3 had a sensitivity of 94%, a specificity of 82%, a positive predictive value of 66%, and a negative predictive value of 97%. Additional analyses showed that the test would still have a negative predictive value of 95% or better up to a cancer/NIFTP prevalence of 44%.

All five false-negative cases in the entire study cohort were intrathyroidal nodules of low stage and without aggressive histology.

Of the 33 false-positive cases, 68% were diagnosed on pathology as Hurthle cell or follicular adenomas, 10% were initially diagnosed by local pathologists as cancer or NIFTP, and 94% harbored clonal oncogenic molecular alterations.

“So, these are not actually hyperplasia; these are true tumors. Probably at least some of them would have the potential to progress,” said Dr. Nikiforov. “I believe that this so-called false-positive rate may not be really false positive. This is a rate of detection of precancerous tumors, not hyperplasia, that still may require surgical excision.”

In this study, “we found very high sensitivity and negative predictive value of ThyroSeq v3, with robust negative predictive value in populations with different disease prevalence,” he concluded. “Robust performance was achieved in many thyroid cancer types, including Hurthle cell cancer.”

All study patients underwent surgery, so it is not clear how the classifier would perform in the context of surveillance, he acknowledged. But the 97% negative predictive value gives confidence for patients having a negative result.

“Those patients very likely can be observed – not necessarily dismissed from medical surveillance, but observed – and could probably avoid surgery,” he said. “If patients have a positive test, it will depend on the type of mutation, because some of them confer a high risk and others confer low risk. So, there may be a spectrum of management based on combination of clinical parameters and molecular testing. But those are more likely to be surgical candidates.”

“This is a study that is desperately needed in this field,” session attendee Bryan McIver, MD, PhD, an endocrinologist and deputy physician-in-chief at the Moffitt Cancer Center in Tampa, said in an interview. “These are very challenging studies to do, because the marketing of these molecular tests has run ahead of a lot of the clinical studies.

Dr. Bryan McIver, an endocrinologist and deputy physician-in-chief at the Moffitt Cancer Center in Tampa

Dr. Bryan McIver

“It’s very hard in the United States, at least, to find patients who are truly naive to molecular testing whom you can take to the operating room,” he explained. “And if you can’t take patients with a negative molecular test to the operating room, then you can’t actually calculate the true sensitivity and specificity of the test, and the whole evaluation of the test starts to become skewed.”

According to Dr. McIver, this study is noteworthy in that it largely fulfills four key criteria: There were no post hoc sample exclusions after unblinding of data, both pathology evaluation and decision to operate were blinded to classifier results, and patients were generally unselected, with little to no prior molecular testing.

“So, we actually have a proper high-quality validation study now available for this new test, the ThyroSeq v3,” he noted. “That sets the bar where it needed to be set a long time ago, and I can’t begin to tell you how excited I am to finally have a test that passed that bar. The fact that it shows a negative predictive value of 97% in this clinical study and a positive predictive value in the mid-60% range means that there is a potential for a clinical utility there that is backed by solid science. In this field, that’s almost unique.”

Next Article: