Gene Expression Classifier Identifies Which 'Indeterminate' Thyroid Nodules are Benign

PHILADELPHIA – A novel gene expression classifier was successful in determining which "indeterminate" thyroid nodules are benign and which were "suspicious" in a prospective, multicenter trial.

Following fine-needle aspiration (FNA) analysis, 15%-30% of thyroid nodules are not clearly benign or malignant. These cytologically indeterminate nodules (including atypia or follicular lesions of undetermined significance [FLUS] and lesions suspicious for follicular/Hurthle cell neoplasm) are often referred for diagnostic surgery, which proves three-fourths to be benign, Dr. Richard Lanman, chief medical officer at Veracyte Inc., said at the annual meeting of the American Association of Clinical Endocrinologists.

Two previous validation studies – one published (J. Clin. Endocrinol. Metab. 2010;95:5296-304) and one presented at the 2010 International Thyroid Congress – showed high sensitivity and a negative predictive value of about 95%, similar to that of a benign cytology diagnosis. The test, Veracyte’s Afirma Thyroid FNA analysis, became commercially available in April of this year.

The test includes 142 genes, which were identified through whole-genome analyses of hundreds of thyroid samples to differentiate benignity from malignancy in indeterminate thyroid FNA samples. "Unlike the candidate malignancy marker approach to look for markers of cancer, our goal is to look for a pattern of benignity across multiple genes. The results are expressed either as benign or suspicious, so it doesn’t make the malignant call," he explained.

Development of the test has been an iterative process, with retraining for each new set of samples that presumably includes additional rare neoplasms. A prospective analysis of the diagnostic performance of the current iteration was conducted at 49 study sites – about one third academic and two-thirds community based – in 26 states. A variety of FNA collection techniques were used but 99% were ultrasound-guided, noted Dr. Lanman, who is co–principal investigator for the study.

Patients, physicians, surgeons, and endocrine pathologists were all blinded to the molecular results. Interestingly, the two endocrine pathologists, whose diagnosis was defined as "truth" for the purposes of the study, changed the surgical pathology diagnosis from malignant to benign, or vice versa, 14% of the time. "This shows the broad multicenter nature of the trial," Dr. Lanman commented.

Of the total 4,812 nodules of 1 cm or greater evaluated, 12% (577) from 532 patients were indeterminate. Of those, 72% (413) were surgically removed, thereby allowing for the endocrine pathologists’ assessment. Following exclusions for prespecified criteria (including inadequate or degraded RNA, excessive study site storage time, and unavailable reference standard), 265 indeterminate nodules remained.

Among the entire group of 265 indeterminate FNAs, 180 were benign and 85 malignant by cytology. The gene expression classifier correctly identified 78 of those 85 as "suspicious," for a sensitivity of 92%. Specificity was 52% and negative predictive value was 93%. There were seven false negatives, Dr. Lanman noted.

Sensitivity was high for each subcategory of cytology diagnosis of atypia/FLUS, follicular neoplasm, and suspicious for malignancy at 90%, 90%, and 94%, respectively. Negative predictive values were 95%, 94%, and 85% for each subcategory, respectively.

Dr. Lanman said that these data support consideration of a conservative observation approach for many patients with indeterminate FNA cytology and a benign gene expression classifier result. However, he noted, "no test is perfect. All of these patients with a gene expression classifier benign result need close follow-up sonographically and clinically."

In an interview, Dr. Rhoda Cobin said that her endocrinology practice has been using the Afirma assay for the last several months. So far they have had one suspicious nodule, but it was borderline in size (9 mm) and turned out benign at surgery, making it a possibly false-positive result. All other samples sent gave benign results and therefore are being followed conservatively.

"The assay is useful in helping to make clinical decisions, but I believe its true value will be determined when it is in wider use," said Dr. Cobin, clinical professor of endocrinology, diabetes, and bone disease at Mount Sinai School of Medicine, New York.

Dr. Lanman is an employee of Veracyte, which funded the study. Dr. Cobin has no disclosures.

PHILADELPHIA – A novel gene expression classifier was successful in determining which "indeterminate" thyroid nodules are benign and which were "suspicious" in a prospective, multicenter trial.

Following fine-needle aspiration (FNA) analysis, 15%-30% of thyroid nodules are not clearly benign or malignant. These cytologically indeterminate nodules (including atypia or follicular lesions of undetermined significance [FLUS] and lesions suspicious for follicular/Hurthle cell neoplasm) are often referred for diagnostic surgery, which proves three-fourths to be benign, Dr. Richard Lanman, chief medical officer at Veracyte Inc., said at the annual meeting of the American Association of Clinical Endocrinologists.

Two previous validation studies – one published (J. Clin. Endocrinol. Metab. 2010;95:5296-304) and one presented at the 2010 International Thyroid Congress – showed high sensitivity and a negative predictive value of about 95%, similar to that of a benign cytology diagnosis. The test, Veracyte’s Afirma Thyroid FNA analysis, became commercially available in April of this year.

The test includes 142 genes, which were identified through whole-genome analyses of hundreds of thyroid samples to differentiate benignity from malignancy in indeterminate thyroid FNA samples. "Unlike the candidate malignancy marker approach to look for markers of cancer, our goal is to look for a pattern of benignity across multiple genes. The results are expressed either as benign or suspicious, so it doesn’t make the malignant call," he explained.

Development of the test has been an iterative process, with retraining for each new set of samples that presumably includes additional rare neoplasms. A prospective analysis of the diagnostic performance of the current iteration was conducted at 49 study sites – about one third academic and two-thirds community based – in 26 states. A variety of FNA collection techniques were used but 99% were ultrasound-guided, noted Dr. Lanman, who is co–principal investigator for the study.

Patients, physicians, surgeons, and endocrine pathologists were all blinded to the molecular results. Interestingly, the two endocrine pathologists, whose diagnosis was defined as "truth" for the purposes of the study, changed the surgical pathology diagnosis from malignant to benign, or vice versa, 14% of the time. "This shows the broad multicenter nature of the trial," Dr. Lanman commented.

Of the total 4,812 nodules of 1 cm or greater evaluated, 12% (577) from 532 patients were indeterminate. Of those, 72% (413) were surgically removed, thereby allowing for the endocrine pathologists’ assessment. Following exclusions for prespecified criteria (including inadequate or degraded RNA, excessive study site storage time, and unavailable reference standard), 265 indeterminate nodules remained.

Among the entire group of 265 indeterminate FNAs, 180 were benign and 85 malignant by cytology. The gene expression classifier correctly identified 78 of those 85 as "suspicious," for a sensitivity of 92%. Specificity was 52% and negative predictive value was 93%. There were seven false negatives, Dr. Lanman noted.

Sensitivity was high for each subcategory of cytology diagnosis of atypia/FLUS, follicular neoplasm, and suspicious for malignancy at 90%, 90%, and 94%, respectively. Negative predictive values were 95%, 94%, and 85% for each subcategory, respectively.

Dr. Lanman said that these data support consideration of a conservative observation approach for many patients with indeterminate FNA cytology and a benign gene expression classifier result. However, he noted, "no test is perfect. All of these patients with a gene expression classifier benign result need close follow-up sonographically and clinically."

In an interview, Dr. Rhoda Cobin said that her endocrinology practice has been using the Afirma assay for the last several months. So far they have had one suspicious nodule, but it was borderline in size (9 mm) and turned out benign at surgery, making it a possibly false-positive result. All other samples sent gave benign results and therefore are being followed conservatively.

"The assay is useful in helping to make clinical decisions, but I believe its true value will be determined when it is in wider use," said Dr. Cobin, clinical professor of endocrinology, diabetes, and bone disease at Mount Sinai School of Medicine, New York.

Dr. Lanman is an employee of Veracyte, which funded the study. Dr. Cobin has no disclosures.

PHILADELPHIA – A novel gene expression classifier was successful in determining which "indeterminate" thyroid nodules are benign and which were "suspicious" in a prospective, multicenter trial.

Following fine-needle aspiration (FNA) analysis, 15%-30% of thyroid nodules are not clearly benign or malignant. These cytologically indeterminate nodules (including atypia or follicular lesions of undetermined significance [FLUS] and lesions suspicious for follicular/Hurthle cell neoplasm) are often referred for diagnostic surgery, which proves three-fourths to be benign, Dr. Richard Lanman, chief medical officer at Veracyte Inc., said at the annual meeting of the American Association of Clinical Endocrinologists.

Two previous validation studies – one published (J. Clin. Endocrinol. Metab. 2010;95:5296-304) and one presented at the 2010 International Thyroid Congress – showed high sensitivity and a negative predictive value of about 95%, similar to that of a benign cytology diagnosis. The test, Veracyte’s Afirma Thyroid FNA analysis, became commercially available in April of this year.

The test includes 142 genes, which were identified through whole-genome analyses of hundreds of thyroid samples to differentiate benignity from malignancy in indeterminate thyroid FNA samples. "Unlike the candidate malignancy marker approach to look for markers of cancer, our goal is to look for a pattern of benignity across multiple genes. The results are expressed either as benign or suspicious, so it doesn’t make the malignant call," he explained.

Development of the test has been an iterative process, with retraining for each new set of samples that presumably includes additional rare neoplasms. A prospective analysis of the diagnostic performance of the current iteration was conducted at 49 study sites – about one third academic and two-thirds community based – in 26 states. A variety of FNA collection techniques were used but 99% were ultrasound-guided, noted Dr. Lanman, who is co–principal investigator for the study.

Patients, physicians, surgeons, and endocrine pathologists were all blinded to the molecular results. Interestingly, the two endocrine pathologists, whose diagnosis was defined as "truth" for the purposes of the study, changed the surgical pathology diagnosis from malignant to benign, or vice versa, 14% of the time. "This shows the broad multicenter nature of the trial," Dr. Lanman commented.

Of the total 4,812 nodules of 1 cm or greater evaluated, 12% (577) from 532 patients were indeterminate. Of those, 72% (413) were surgically removed, thereby allowing for the endocrine pathologists’ assessment. Following exclusions for prespecified criteria (including inadequate or degraded RNA, excessive study site storage time, and unavailable reference standard), 265 indeterminate nodules remained.

Among the entire group of 265 indeterminate FNAs, 180 were benign and 85 malignant by cytology. The gene expression classifier correctly identified 78 of those 85 as "suspicious," for a sensitivity of 92%. Specificity was 52% and negative predictive value was 93%. There were seven false negatives, Dr. Lanman noted.

Sensitivity was high for each subcategory of cytology diagnosis of atypia/FLUS, follicular neoplasm, and suspicious for malignancy at 90%, 90%, and 94%, respectively. Negative predictive values were 95%, 94%, and 85% for each subcategory, respectively.

Dr. Lanman said that these data support consideration of a conservative observation approach for many patients with indeterminate FNA cytology and a benign gene expression classifier result. However, he noted, "no test is perfect. All of these patients with a gene expression classifier benign result need close follow-up sonographically and clinically."

In an interview, Dr. Rhoda Cobin said that her endocrinology practice has been using the Afirma assay for the last several months. So far they have had one suspicious nodule, but it was borderline in size (9 mm) and turned out benign at surgery, making it a possibly false-positive result. All other samples sent gave benign results and therefore are being followed conservatively.

"The assay is useful in helping to make clinical decisions, but I believe its true value will be determined when it is in wider use," said Dr. Cobin, clinical professor of endocrinology, diabetes, and bone disease at Mount Sinai School of Medicine, New York.

Dr. Lanman is an employee of Veracyte, which funded the study. Dr. Cobin has no disclosures.