We read with interest the commentary by Dr. Faquin concerning his findings regarding gene-expression classifiers, and share some dismays as well as highlight our standpoints on this debatable topic.
The article by Alexander et al focused on the use of a gene-expression classifier to improve the preoperative risk assessment for the indeterminate category of thyroid cytology, leading to a 95% negative predictive value (NPV) for a classification of atypia of unknown significance and 94% for a classification of follicular neoplasm, with a 62% false-positive rate for the former category. The data indicated a high NPV and sensitivity versus “classic” molecular applications, which are largely the most diffuse assays. The Afirma Thyroid FNA Analysis (Veracyte, South San Francisco, Calif) proposes an approach resulting in apparently simple findings in super-specialized centers with expensive economical knock-on effects.
In agreement with Faquin, we noted that Alexander et al underestimated the “great power” of morphological interpretation supported by the invaluable opportunity to perform repeat fine-needle aspiration biopsy. We agree that the controversial “gray zone” of indeterminate diagnoses has induced many “expert teams” to accept the challenge of new “ancillary” approaches for better diagnostic accuracy in these categories.[3, 4]
In our experience, excellent sensitivity and NPV could be replaced by the easier and more cost-effective use of immunocytochemical (ICC) panels. We highlighted the role of ICC panels for indeterminate lesions and demonstrated a 77% NPV and 91% sensitivity, especially with liquid-based cytology. In this category, ICC sorts out “high- and low-risk” lesions, driving only the 3.2% of false-negative cases to molecular detection. Our unpublished data on the further molecular application on an ICC-negative panel demonstrated a positivity rate of 13.3% for BRAF and N-Ras.
We acknowledge that ICC is an indicator of malignant risk without any specific definition of the nature of the lesion but our results regarding the role of molecular testing such as BRAF and N-Ras have nonetheless reached a high specificity but a less lower sensitivity, most likely related to the evidence that the majority of cancers in this category are follicular variants of papillary carcinoma with a very low incidence of BRAF mutations.
In ruling out a malignant neoplasm, we encourage the first-line application of ICC as a first screening test for indeterminate and follicular neoplasm lesions with a sequential detection of molecular panels. Therefore, we would suggest promoting the use of a “high cost” gene-classifier as a second-level test when ICC or “classic molecular analyses” do not achieve a significant or conclusive result.