NEW YORK – GeneproDx, a Santiago, Chile-based startup, is developing a multiplex quantitative PCR assay that it believes will help patients at risk of thyroid cancer avoid unnecessary surgery.
Publishing a pair of validation studies earlier this month on the firm’s ThyroidPrint assay, US and Chilean researchers showed they could successfully predict benign thyroid nodules with a sensitivity of 90 to 91 percent and a specificity of 87 to 89 percent.
The company now plans to launch an in vitro diagnostic version of the 10-gene ThyroidPrint assay in South America and Europe after receiving regulatory clearance by the second quarter of 2021.
Clinicians usually determine the risk of a patient's thyroid nodule developing into thyroid cancer by performing cytology on a fine needle aspirate sample using the six-point Bethesda Classification scale. They classify most samples as non-diagnostic or benign (I or II) and having a very low risk of cancer, or highly suspicious or visibly malignant (V or VI) and requiring removal of most if not all the patient's thyroid.
However, clinicians have struggled to determine the status of some samples that exist between these categories and often must perform repeated and invasive fine needle aspirates to monitor nodule development.
"When you've got Bethesda III and IV 'middle of the road' nodules, most of those end up being benign, but on the other hand, 20 to 25 percent of them have cancer, and most patients don't want to risk that chance," explained Mark Zafereo, first author on the recently published validation study and an associate professor of head and neck surgery at the University of Texas MD Anderson Cancer Center. "However, thyroid cancer surgery is not without risk, as you're putting three out of four [patients] through unnecessary surgery that could affect their quality of life."
GeneproDx spun out of the Catholic University of Chile in 2014 led by Hernán González, an associate professor of surgical oncology-endocrine surgery at the university. The company has exclusively licensed patents related to the technology from the university and has raised a total of $6 million in seed funding, with support from angel investors and Chilean venture capital firm Fondo Alerce.
The company's ThyroidPrint genetic classifier is intended to more accurately and quickly establish if a patient with indeterminate thyroid nodules is at risk of developing thyroid cancer. The firm has offered the assay as a laboratory-developed test out of its CLIA-certified, CAP-accredited lab in Santiago since early 2019. After receiving a patient's fine needle aspirate sample from a clinician, GeneproDx technicians test it using the ThyroidPrint assay, which runs on a Qiagen Rotor-Gene Q MDx qPCR instrument.
While GeneproDx currently uses Qiagen’s tool for the ThyroidPrint platform, González noted that the firm may switch to another instrument in the future, depending on the format of its IVD assay.
The team then processes the qPCR data using a proprietary algorithm that the group trained via a machine learning iterative process, which provides a score that predicts whether a sample is benign or malignant.
"The test can change the probability of malignancy from 25 percent to anywhere from 4 to 75 percent, basically addressing the clinical uncertainty that physicians face with patients," González explained.
González said that ThyroidPrint only requires about three to four days to provide clinically actionable results.
In the validation study published in Thyroid earlier this month, González's team examined the genetic classifier's performance in two independent, ethnically diverse, prospective, multicenter trials called TGCT-1 and TGCT-2. González performed the initial trial in Chile, while Zafereo led the subsequent US trial.
"Hernan had initially performed the first study in Chile, where he'd seen good data from that cohort," Zafereo noted. "However, he needed to validate the study in other populations and in major academic centers in the US."
The team collected a total of 4,061 fine-needle aspirates from 15 institutions between Chile (TGCT-1) and the US (TGCT-2), identifying 897 indeterminate samples between the cohorts.
The group extracted RNA from the indeterminate samples, followed by cDNA synthesis and sample cold storage. The researchers also collected patient data including demographics and ultrasound thyroid nodule characteristics. In each trial, a central expert pathologist provided surgical pathology reports, which were not shared with the patient, pathologist, or treating physicians.
After a four-month follow up, 441 patients received surgical resection – 181 Bethesda III and 252 Bethesda IV – and ThyroidPrint successfully provided informative valid classifier scores in 319 patients. After randomly assigning 49 samples to a training set, the team developed a validation set of 270 cases.
To examine the classifier performance, the team divided surgical pathologic diagnoses into non-surgical or surgical groups. The ThyroidPrint classifier provided a score for each sample that was categorized into "benign" or "suspicious for malignancy."
After the group independently uploaded both the de-identified pathology reports and classifier scores to an electronic capture system, an independent third party then matched each report to their corresponding classifier result.
The researchers saw significant differences between the two cohorts, including clinical site patient source, ethnic diversity, and tumor size. Only 43 percent of samples came from academic sources from the Chilean cohort, while 97 percent of the US cohort came from academic sources. González noted that this was because his team worked with medical community centers in the Chilean cohort. Tumor size averaged about 1.7 cm and 2.5 cm within TGCT-1 and TGCT-2 respectively.
At the same time, the two independent groups did not identify any differences in the histopathological profiles between the cohorts.
The researchers found that ThyroidPrint had a clinical sensitivity of 90 percent and specificity of 89 percent in TGCT-1, while producing clinical sensitivity of 91 percent and specificity of 87 percent in TGCT-2. The group also saw that ThyroidPrint had a negative predictive value of 96 percent in TGCT-1 and 94 percent in TGCT-2.
González highlighted that ThyroidPrint showed high accuracy in a wide spectrum of thyroid conditions that normally require surgery, including papillary thyroid carcinomas, follicular carcinomas, Hurthle cell carcinomas, and noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP).
However, the team noted that the study had multiple limitations related to patient enrollment and selection bias. In the initial phase of patient enrollment, the sample RNA yield failure reached about 14 percent. The group also found that the rate of indeterminate cases undergoing survey in the US cohort was 65 percent compared to 43 percent in the Chilean cohort, which could possibly introduce selection bias.
While receiving research funding for the validation study, Zafereo does not have any financial ties to GeneproDx.
González noted that GeneproDx is now raising additional funding to develop an IVD version of the assay, which the firm envisions offering to hospitals and reference labs to improve turnaround times for results. While the firm currently offers a research-use-only version of the assay in multiple Latin American countries, González aims to offer the test as a distributable kit in other regions such as Europe.
GeneproDx is therefore seeking different forms of regulatory clearance in multiple regions, such as CE-IVD marking for Europe. The firm would like to eventually tackle the US market, but González declined to provide a potential timeline for US entry.
"We want to globally distribute the assay, as beyond the US, there is no similar test current being offered," González said, who expects to reach this goal within the next 18 to 24 months.
"Since firms like Veracyte has a big sector of the [US] market, it doesn’t seem to make sense right now to compete at the same level for a small company like ours, but once [the test] is an IVD service, that’s the moment we will be more aggressive on changing our commercial strategy," González explained. "A centralized service in near term outside the US is a possibility, but the US market will need to be addressed in a different way."
GeneproDx has also begun a third multicenter trial, taking place in the US, Chile, Brazil, the UK, and Spain to increase the amount of evidence for ThyroidPrint in an IVD format. Partnering with academic centers including the MD Anderson, the University of Edinburgh, and the Guy's and St. Thomas National Health Service Foundation, GeneproDx will collect about 250 new patient samples and anticipates completing the project by early 2021.
González said that GeneproDx aims to use the trial's results to eventually provide other countries access to the assay.
"We know [that] the test works well in a broad spectrum of ethnicities and clinical settings, but we want to test its performance in other countries," González explained
In the US, Veracyte offers its Afirma genomic sequencing classifier, which uses RNA whole-transcriptome sequencing and machine learning to identify benign and indeterminate nodules. The firm previously received coverage for the assay from Anthem in 2017 and now plans to reach a global audience by transitioning its diagnostic and prognostic cancer assays onto NanoString's nCounter platform.
Sonic Healthcare currently licenses the University of Pittsburgh Medical Center's ThyroSeq assay, which uses next-generation sequencing to analyze DNA and RNA collected via fine needle aspiration to distinguish between benign and malignant thyroid nodules.
In Brazil, startup Onkos Molecular Diagnostics seeks to commercialize its mir-THYpe thyroid molecular classifier, which is also a qPCR gene expression test that helps determine 11 microRNAs to help determine if nodules are benign or likely cancerous.
González highlighted that ThyroidPrint stands out from currently available assays on the market because of its improved specificity and need for fewer biomarkers. By only using 10 biomarkers, González believes that ThyroidPrint will be easier to migrate into an IVD kit for underserved global markets.
Veracyte CEO Bonnie Anderson said in an email that focusing on a specific set number of genes to develop a molecular assay is very limiting for determining thyroid cancer risk in a patient's sample.
"It's similar to looking for your lost keys under a lamp post because that's the only place you can see," Anderson argued. "Our Afirma GSC, [however], is developed using RNA whole-transcriptome sequencing and machine learning, which means we scour the entire genome and let the algorithm determine which features, including RNA expression, variants, fusions, and copy numbers … are important to distinguish benign from likely cancerous nodules."
González said that the GeneproDx plans to launch ThyroidPrint next month in Brazil, with molecular diagnostic firm DASA offering the test as a centralized service to the country.