Identification of H\xfcrthle cell cancers: solving a clinical challenge with genomic sequencing and a trio of machine learning algorithms

BackgroundAnalysis of cytologically indeterminate thyroid nodules with Afirma Gene Expression Classifier (GEC) and Genomic Sequencing Classifier (GSC) can reduce surgical rate and increase malignancy rate of surgically resected indeterminate nodules.MethodsRetrospective cohort analysis of all adults with cytologically indeterminate thyroid nodules from January 2013 through December 2019. We compared surgical and malignancy rates of those without molecular testing to those with GEC or GSC, analyzed test performance between GEC and GSC, and identified variables associated with molecular testing.Results468 indeterminate thyroid nodules were included. No molecular testing was performed in 273, 71 had GEC, and 124 had GSC testing. Surgical rate was 68% in the group without molecular testing, 59% in GEC, and 40% in GSC. Malignancy rate was 20% with no molecular testing, 22% in GEC, and 39% in GSC (P = 0.022). GEC benign call rate (BCR) was 46%; sensitivity, 100%; specificity, 61%; and positive predictive value (PPV), 28%. GSC BCR was 60%; sensitivity, 94%; specificity, 76%; and PPV, 41%. Those with no molecular testing had larger nodule size, preoperative growth of nodules, and constrictive symptoms and those who underwent surgery in the no molecular testing group had higher body mass index, constrictive symptoms, higher Thyroid Imaging Reporting and Data System and Bethesda classifications. Type of provider was also associated with the decision to undergo surgery.ConclusionImplementation of GEC showed no effect on surgical or malignancy rate, but GSC resulted in significantly lower surgical and higher malignancy rates. This study provides insight into the factors that affect the real-world use of these molecular markers preoperatively in indeterminate thyroid nodules.

Background: For thyroid nodules with indeterminate cytology, the Afirma Gene Expression Classifier (GEC) identified benign nodules to reduce diagnostic surgery, though many nodules classified as suspicious still proved histopathologically benign. The current Afirma Genomic Sequencing Classifier (GSC) demonstrates improved specificity, suggesting more nodules will have a benign result (benign call rate [BCR]), but independent data are needed to confirm this in clinical practice. Methods: Retrospective analysis was performed of all Bethesda III or IV cytology thyroid nodules ≥1 cm tested with GEC (between January 1, 2011, and July 19, 2017) or GSC (between July 20, 2017, and August 27, 2018) at the authors' institution. Afirma testing was not performed reflectively for all nodules with Bethesda III or IV cytology, but rather was applied based on physician-patient decision making. Demographic, sonographic, and cytologic data were collected. The BCR for GEC- versus GSC-tested nodules was compared and further stratified by Bethesda classifications. Results: The study evaluated 600 nodules in 563 patients tested with either GEC (n = 486) or GSC (n = 114). The BCR was 233/486 (47.9%) for the GEC compared to 75/114 (65.8%) for the GSC (p = 0.0006). Hürthle-cell cytology was present in 99/486 (20.4%) nodules in the GEC group compared to 31/114 (27.2%) nodules in the GSC group (p = 0.28). The GSC BCR was significantly higher than the GEC BCR for Bethesda III nodules characterized by Hürthle cells (p = 0.006), but the BCRs were similar for nodules with architectural or cytologic atypia. In Bethesda IV nodules suspicious for follicular neoplasm, BCR for the GEC and GSC were similar (p = 0.68), but for cytology suspicious for Hürthle-cell neoplasm, the GSC BCR was 68.2% (15/22) compared to the GEC BCR of 16.4% (10/61; p < 0.0001). Positive predictive value in resected nodules with a suspicious result was 16/32 (50%) for GSC nodules and 75/221 (33.9%) for GEC nodules (p = 0.1). Conclusions: The higher BCR for the GSC compared to the GEC for indeterminate thyroid nodules, predominantly among nodules with Hürthle-cell cytology, will likely lead to further reduction in surgical management.

Afirma gene expression classifier (GEC) is an adjunct to thyroid fine needle aspiration shown to improve pre-operative risk assessment and reduce unnecessary surgery of indeterminate thyroid nodules. Genomic sequencing classifier (GSC) is a newer version aiming to improve specificity and positive predictive value (PPV) of Afirma testing. There are limited studies comparing GSC vs GEC. This study was undertaken to compare these classifiers in terms of diagnostic performance and effect on clinical management of indeterminate thyroid nodules. The study cohort consisted of patients with thyroid nodules that had a recurrent cytologic diagnosis of atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS) and were tested by either GEC or GSC. Patient demographics, nodule size, and clinical follow-up were recorded. Benign call rate (BCR) of Afirma testing, rate of subsequent surgery (RSS), rate of histology-confirmed malignancy (RHM), as well as diagnostic sensitivity, specificity, PPV, negative predicative value (NPV), and accuracy were calculated and compared between GSC and GEC cohorts. Among 264 AUS/FLUS thyroid nodules, 127 and 137 were tested with GEC and GSC, respectively. Compared to GEC, GSC demonstrated increased BCR (77.3% vs 52%), decreased RSS (31.4% vs 51.2%), greater RHM (29% vs 9.8%) associated with a suspicious Afirma result, as well as improved specificity (82.8% vs 54.5%), PPV (29% vs 9.8%), and diagnostic accuracy (83.9% vs 56.7%), while maintaining high sensitivity and NPV. Afirma GSC substantially improved BCR, RSS, RHM, and diagnostic performance, enhancing appropriate triage and thereby helped avoid unnecessary surgery in AUS/FLUS thyroid nodules.

Use of next-generation sequencing of RNA and machine learning algorithms can classify the risk of malignancy in cytologically indeterminate thyroid nodules to limit unnecessary diagnostic surgery. To measure the performance of a genomic sequencing classifier for cytologically indeterminate thyroid nodules. A blinded validation study was conducted on a set of cytologically indeterminate thyroid nodules collected by fine-needle aspiration biopsy between June 2009 and December 2010 from 49 academic and community centers in the United States. All patients underwent surgery without genomic information and were assigned a histopathology diagnosis by an expert panel blinded to all genomic information. There were 210 potentially eligible thyroid biopsy samples with Bethesda III or IV indeterminate cytopathology that constituted a cohort previously used to validate the gene expression classifier. Of these, 191 samples (91.0%) had adequate residual RNA for validation of the genomic sequencing classifier. Algorithm development and independent validation occurred between August 2016 and May 2017. Thyroid nodule surgical histopathology diagnosis by an expert panel blinded to all genomic data. The primary end point was measurement of genomic sequencing classifier sensitivity, specificity, and negative and positive predictive values in biopsies from Bethesda III and IV nodules. The secondary end point was measurement of classifier performance in biopsies from Bethesda II, V, and VI nodules. Of the 183 included patients, 142 (77.6%) were women, and the mean (range) age was 51.7 (22.0-85.0) years. The genomic sequencing classifier had a sensitivity of 91% (95% CI, 79-98) and a specificity of 68% (95% CI, 60-76). At 24% cancer prevalence, the negative predictive value was 96% (95% CI, 90-99) and the positive predictive value was 47% (95% CI, 36-58). The genomic sequencing classifier demonstrates high sensitivity and accuracy for identifying benign nodules. Its 36% increase in specificity compared with the gene expression classifier potentially increases the number of patients with benign nodules who can safely avoid unnecessary diagnostic surgery.

In Brief

The use of fine-needle aspiration (FNA) to triage thyroid nodules has resulted in a significant reduction in thyroid surgery. However, approximately one-third of FNA specimens fall into the "indeterminate" category. The Afirma gene expression classifier (GEC) has been used to identify benign nodules with a high sensitivity and negative predictive value. However, the specificity and positive predictive value of the "suspicious" category are low. The updated Afirma genomic sequencing classifier (GSC) has been reported to demonstrate increased specificity while maintaining a high sensitivity and negative predictive value. The authors retrospectively investigated 272 indeterminate thyroid FNA specimens (Bethesda categories III and IV) from nodules measuring >1cm using the Afirma GEC or GSC tests (July 2012-January 2019). Of the 194 nodules tested using the Afirma GEC, a benign result was obtained in 88 cases (45.4%). In comparison, 52 of 78 FNA samples (66.7%) tested using GSC yielded a benign result (P=.002). In the GEC group, there were 31 cases with oncocytic cytology, 5 of which (16.1%) were benign on Afirma and 26 of which (83.9%) were suspicious on Afirma. In contrast, in the GSC group, there were 10 cases with oncocytic cytology, 8 of which (80%) were benign on Afirma and only 2 of which (20%) were found to be suspicious on Afirma (P<.001). The positive predictive value of the GSC group (57.1%) was higher than that of the GEC group (36.7%); however, there was no statistical significance noted (P=.15). A larger percentage of indeterminate thyroid FNA specimens were classified as benign using the Afirma GSC compared with the Afirma GEC, especially among samples with oncocytic features. The Afirma GSC appears to have a higher benign call rate compared with the Afirma GEC.

ROLE OF NEXT GENERATION SEQUENCING WITH PERCEPTA BRUSHING IN RECLAFFYING LUNG NODULE RISK AFTER A NON-DIAGNSOTIC BRONCHOSCOPY

Background: Thyroid nodules are a very common often incidental finding on physical examination or imaging. Of those who undergo fine needle aspiration, cytology is indeterminate in up to 15%. Molecular testing is increasingly being used to help identify which nodules may be high risk for malignancy and guide management with regard to clinical follow-up or surgical intervention. Recently there has been an increase in publication of independent studies assessing the performance of these molecular tests and comparing "real-world" data with the validation studies. Methods: This retrospective study identified all thyroid nodules at our institution that had Afirma gene expression classifier (GEC), genomic sequencing classifier (GSC), or Thyroseq v3 molecular testing from January 2014 to January 2020 and compared measurements of test performance between them at our institution, and then with the original validation studies and other published institutional data. Results: Overall, the benign call rate was highest in the Afirma GSC group (78%) compared with the GEC group (60%) and Thyroseq group (66%). Surgical histopathology revealed malignancy in 6 of 31of biopsied nodules in the GEC group, 8 of 13 in the GSC group, and 3 of 16 in the Thyroseq v3 group. Based on our data, the GSC specificity (73.7%) and positive predictive value (PPV) (61.5%) were higher than the GEC specificity (60.4%) and PPV (22.2%) as well as Thyroseq v3 specificity (55.2%) and PPV (18.8%). Conclusions: From our short-term institutional experience, we found that the GSC classified more cytologically indeterminate nodules as benign compared with the Afirma GEC, and had improved specificity and PPV, which is similar to the validation study and other institutions' reported experiences. We also found that the Thyroseq v3 was similar to the Afirma GEC in terms of specificity and PPV, both of which are much lower than the validation studies.

The Gene Expression Classifier (GEC) and Genomic Sequencing Classifier (GSC) were developed to improve risk stratification of indeterminate nodules. Our aim was to assess the clinical utility in a European population with restrictive diagnostic workup. Clinical utility of the GEC was assessed in a prospective multicenter cohort of 68 indeterminate nodules. Diagnostic surgical rates for Bethesda III and IV nodules were compared to a historical cohort of 171 indeterminate nodules. Samples were post hoc tested with the GSC. The GEC classified 26% as benign. Surgical rates between the prospective and historical cohort did not differ (72.1% vs. 76.6%). The GSC classified 59% as benign, but misclassified six malignant lesions as benign. Implementation of GEC in management of indeterminate nodules in a European country with restrictive diagnostic workup is currently not supported, especially in oncocytic nodules. Prospective studies with the GSC in European countries are needed to determine the clinical utility.

8549 Background: Current guidelines recommend that patients who have lung nodules with high risk of malignancy (ROM) ( &gt; 65%) should undergo surgical and other ablative therapies. However, prior studies have shown that clinicians may opt for more conservative management in these high-risk patients. Percepta Genomic Sequencing Classifier (GSC), a RNA-seq based classifier derived from bronchial epithelial cells to assess risk of lung cancer, was designed to risk stratify lung nodules by both down classifying ROM as a “rule -out“ test with high sensitivity as well as up-classifying ROM as a “rule- in” test with high specificity for malignancy. This study assesses the impact of up-classification of high ROM to very high- risk (ROM &gt; 90%) by Percepta GSC in increasing the number of ablative therapies recommended for high-risk lung nodules. Methods: This prospective randomized decision impact survey included 37 patients from the AEGIS I/ II cohorts and the Percepta Registry who were undergoing work up of a lung nodule and had a high ROM that was up-classified to very high ROM by Percepta GSC. 97 physicians assessed 10 randomly assigned patient cases. They then responded to a survey designed to test the hypothesis that including a Percepta GSC result will increase the recommendation for surgical or other ablative therapy in very high- risk patients as well as their level of confidence of this recommendation. Physicians were first presented with the patient’s clinical information without Percepta GSC and then with Percepta GSC. Results: 97 physicians provided a total of 682 evaluations of 37 patients. In this study, the recommendation for surgical or other ablative therapy increased from 19/341 (5.6%) prior to the Percepta GSC result to 157/341 (46%) after the Percepta GSC result (odds ratio of 4.76, p-value &lt; 0.001). The number of extremely confident recommendations increased from 72/341 (21%) without Percepta GSC to 106/341 (31%) with Percepta GSC. Significantly more physicians had increased confidence in their recommended next step post-Percepta GSC when collapsing the confidence level responses into increased confidence (n = 93) and decreased confidence (n = 44) (p-value = 0.002). Conclusions: Percepta GSC had a quantifiable impact on clinical decision making. It increased the number of surgical and other ablative therapies recommended when patients were re-classified from high to very high- risk of lung cancer with a higher confidence in the recommended next step. By up-classifying nodules from high to very high ROM, Percepta GSC will improve the likelihood and timeliness of appropriate therapies and assist clinicians more effectively manage patients to improve patient outcomes.

BackgroundIncidental and screening-identified lung nodules are common, and a bronchoscopic evaluation is frequently nondiagnostic. The Percepta Genomic Sequencing Classifier (GSC) is a genomic classifier developed in current and former smokers which can be used for further risk stratification in these patients. Percepta GSC has the capability of up-classifying patients with a pre-bronchoscopy risk that is high (> 60%) to “very high risk” with a positive predictive value of 91.5%. This prospective, randomized decision impact survey was designed to test the hypothesis that an up-classification of risk of malignancy from high to very high will increase the rate of referral for surgical or ablative therapy without additional intervening procedures while increasing physician confidence.MethodsData were collected from 37 cases from the Percepta GSC validation cohort in which the pre-bronchoscopy risk of malignancy was high (> 60%), the bronchoscopy was nondiagnostic, and the patient was up-classified to very high risk by Percepta GSC. The cases were randomly presented to U.S pulmonologists in three formats: a pre-post cohort where each case is presented initially without and then with a GSG result, and two independent cohorts where each case is presented either with or without with a GSC result. Physicians were surveyed with respect to subsequent management steps and confidence in that decision.ResultsOne hundred and one survey takers provided a total of 1341 evaluations of the 37 patient cases across the three different cohorts. The rate of recommendation for surgical resection was significantly higher in the independent cohort with a GSC result compared to the independent cohort without a GSC result (45% vs. 17%, p < 0.001) In the pre-post cross-over cohort, the rate increased from 17 to 56% (p < 0.001) following the review of the GSC result. A GSC up-classification from high to very high risk of malignancy increased Pulmonologists’ confidence in decision-making following a nondiagnostic bronchoscopy.ConclusionsUse of the Percepta GSC classifier will allow more patients with early lung cancer to proceed more rapidly to potentially curative therapy while decreasing unnecessary intervening diagnostic procedures following a nondiagnostic bronchoscopy.

6083 Background: Receptor tyrosine kinase (RTK) fusions may be targeted by small molecule inhibitors to treat various advanced tumors, including thyroid cancer. Clinical trials have studied selective inhibitors of ALK, BRAF, NTRK and RET, leading to several FDA-approved therapies. The Afirma Genomic Sequencing Classifier (GSC) classifies cytologically indeterminate thyroid nodules as molecularly benign or suspicious. The Xpression Atlas reports 905 genomic variants and 235 fusion pairs on GSC Suspicious, Suspicious for Malignancy (SFM), and Malignant FNA samples at the time of diagnosis. Here we report the prevalence of these fusion genes in real-world clinical practice. Methods: We analyzed anonymized data from 50,644 consecutive Bethesda III-VI nodule FNA samples submitted to the Veracyte CLIA laboratory for molecular testing using whole transcriptome RNA sequencing (RNA-Seq). Gene pairs are listed alphabetically. Results: 32,080 Bethesda III/IV nodules were classified as GSC Benign and 278 were Parathyroid Classifier positive. No ALK, BRAF, NTRK1/3, or RET fusions were identified among these samples. Among 16,594 Bethesda III/IV GSC Suspicious FNAs, 3% (n = 529) were positive for ALK, BRAF, NTRK1/3 or RET fusions. Among the 1,692 Bethesda V/VI FNAs, the proportion of positive nodules was 8% (n = 135). Among these combined cohorts of Bethesda III/IV GSC Suspicious and Bethesda V/VI, the most common gene fusions observed for each of the 5 studied RTK genes was: ETV6/NTRK3 (n = 164, 72% of NTRK3 fusions), CCDC6/RET (n = 104, 55% of RET), BRAF/SND1 (n = 32, 20% of BRAF), ALK/STRN (n = 20, 37% of ALK), and NTRK1/TPM3 (n = 14, 50% of NTRK1). BRAF showed the highest diversity of fusions, with 80 gene partners. Different gene partners with RET, ALK, NTRK1, and NTRK3 numbered 25, 11, 9, and 5 , respectively . Conclusions: Whole-transcriptome RNA-seq on small sample thyroid FNA specimens can identify clinically relevant ALK, BRAF, NTRK, and RET fusions across Bethesda categories. The prevalence ranges from 3% in Bethesda III/IV Afirma GSC Suspicious specimens to 8% among Bethesda V/VI specimens. Future studies need to determine if detection of precision medicine candidates by pre-operative FNA can optimize initial treatment, predict response to treatment, and prioritize selective targeted therapy should systemic treatment be needed.[Table: see text]

BackgroundBronchoscopy for suspected lung cancer has low diagnostic sensitivity, rendering many inconclusive results. The Bronchial Genomic Classifier (BGC) was developed to help with patient management by identifying those with low risk of lung cancer when bronchoscopy is inconclusive. The BGC was trained and validated on patients in the Airway Epithelial Gene Expression in the Diagnosis of Lung Cancer (AEGIS) trials. A modern patient cohort, the BGC Registry, showed differences in key clinical factors from the AEGIS cohorts, with less smoking history, smaller nodules and older age. Additionally, we discovered interfering factors (inhaled medication and sample collection timing) that impacted gene expressions and potentially disguised genomic cancer signals.MethodsIn this study, we leveraged multiple cohorts and next generation sequencing technology to develop a robust Genomic Sequencing Classifier (GSC). To address demographic composition shift and interfering factors, we synergized three algorithmic strategies: 1) ensemble of clinical dominant and genomic dominant models; 2) development of hierarchical regression models where the main effects from clinical variables were regressed out prior to the genomic impact being fitted in the model; and 3) targeted placement of genomic and clinical interaction terms to stabilize the effect of interfering factors. The final GSC model uses 1232 genes and four clinical covariates – age, pack-years, inhaled medication use, and specimen collection timing.ResultsIn the validation set (N = 412), the GSC down-classified low and intermediate pre-test risk subjects to very low and low post-test risk with a specificity of 45% (95% CI 37–53%) and a sensitivity of 91% (95%CI 81–97%), resulting in a negative predictive value of 95% (95% CI 89–98%). Twelve percent of intermediate pre-test risk subjects were up-classified to high post-test risk with a positive predictive value of 65% (95%CI 44–82%), and 27% of high pre-test risk subjects were up-classified to very high post-test risk with a positive predictive value of 91% (95% CI 78–97%).ConclusionsThe GSC overcame the impact of interfering factors and achieved consistent performance across multiple cohorts. It demonstrated diagnostic accuracy in both down- and up-classification of cancer risk, providing physicians actionable information for many patients with inconclusive bronchoscopy.

BackgroundBronchoscopy is a common procedure used for evaluation of suspicious lung nodules, but the low diagnostic sensitivity of bronchoscopy often results in inconclusive results and delays in treatment. Percepta Genomic Sequencing Classifier (GSC) was developed to assist with patient management in cases where bronchoscopy is inconclusive. Studies have shown that exposure to tobacco smoke alters gene expression in airway epithelial cells in a way that indicates an increased risk of developing lung cancer. Percepta GSC leverages this idea of a molecular “field of injury” from smoking and was developed using RNA sequencing data generated from lung bronchial brushings of the upper airway. A Percepta GSC score is calculated from an ensemble of machine learning algorithms utilizing clinical and genomic features and is used to refine a patient’s risk stratification.MethodsThe objective of the analysis described and reported here is to validate the analytical performance of Percepta GSC. Analytical performance studies characterized the sensitivity of Percepta GSC test results to input RNA quantity, the potentially interfering agents of blood and genomic DNA, and the reproducibility of test results within and between processing runs and between laboratories.ResultsVarying the amount of input RNA into the assay across a nominal range had no significant impact on Percepta GSC classifier results. Bronchial brushing RNA contaminated with up to 10% genomic DNA by nucleic acid mass also showed no significant difference on classifier results. The addition of blood RNA, a potential contaminant in the bronchial brushing sample, caused no change to classifier results at up to 11% contamination by RNA proportion. Percepta GSC scores were reproducible between runs, within runs, and between laboratories, varying within less than 4% of the total score range (standard deviation of 0.169 for scores on 4.57 scale).ConclusionsThe analytical sensitivity, analytical specificity, and reproducibility of Percepta GSC laboratory results were successfully demonstrated under conditions of expected day to day variation in testing. Percepta GSC test results are analytically robust and suitable for routine clinical use.

IMPACT OF PERCEPTA BRONCHIAL GENOMIC CLASSIFIER ON LUNG NODULE MANAGEMENT AT AN ACADEMIC MEDICAL CENTER