NEW YORK – A team led by researchers at the University of Texas and Baylor College of Medicine have developed a mass spectrometry-based test for diagnosing thyroid cancer.
Detailed in a study published last week in Proceedings of the National Academy of Sciences, the approach uses desorption electrospray ionization mass spectrometry (DESI-MS) imaging to analyze fine-needle aspiration (FNA) biopsies and could help evaluate cases where traditional cytology is inconclusive, said James Sulibark, associate professor and chief, endocrine surgery at BCM and author on the paper.
Doctors typically test potentially malignant thyroid nodules by taking FNA biopsies and analyzing them using cytology. In roughly 20 percent of these cases, however, the analysis is inconclusive. These patients often then receive thyroid surgery, and in the majority of cases the nodule is ultimately determined to be benign. Sulibark said that in addition to relatively rare but nonetheless real complications like scarring and voice changes, between 30 and 40 percent of patients who undergo surgery for removal of a nodule will require thyroid hormone supplementation for the rest of their life.
Sulibark noted that genetic assays including Veracyte's Afirma test and the ThyroSeq test offered by CBLPath and the University of Pittsburgh Medical Center are available to help doctors evaluate indeterminate FNA biopsies, but he said that these tests left room for improvement in terms of accuracy. According to the study authors, the Afirma test has specificity of 68 percent and sensitivity of 91 percent, while the ThyroSeq assay has specificity of 82 percent and sensitivity of 94 percent, making both effective tests for ruling out cancer but with relatively low specificity.
Additionally, Sulibark said these tests typically have turnaround times of two to four weeks, which he noted can increase patient anxiety and the cost of care.
Sulibark and his colleagues set out to tackle the problem using imaging mass spec, generating molecular profiles of indeterminate thyroid FNA biopsies that they used to distinguish between benign and malignant cases. Using DESI-MS they analyzed 178 tissue samples to generate profiles of benign nodules, follicular adenoma (FTA), malignant follicular carcinoma (FTC), and papillary carcinoma (PTC). They built classifiers based on this data that they then tested on a set of 69 FNA biopsies from 57 patients — 21 from benign nodules, 19 from FTA, 28 from FTC, and 1 from FTA. Of the nodules in the test set, 24 had indeterminant cytology.
The researchers built two classifiers, one that performed with 93 percent accuracy (100 percent sensitivity and 88 percent specificity) and one that had 89 percent accuracy (96 percent sensitivity and 91 percent specificity). While based on small sample sets, Sulibark said that the performance approached that of cytology, which has an accuracy of around 95 percent in the 80 percent of thyroid nodules it is able to analyze conclusively.
"If we can get to 95 percent then we'll be at the time same accuracy rate as current cytological techniques," he said.
The classifier relies most heavily on the lipid profiles of the nodules, which Sulibark said was not unexpected based on previous work in the area.
"In a prior publication we identified a lipid molecular that was 95 percent-plus predictive of Hurthle cell carcinoma, so we think [lipid analysis] remains an area for further discovery in other thyroid cancer types," he said.
The researchers used DESI imaging mass spec for their analysis, which Sulibark said has the advantage of a simple workflow and being non-destructive, meaning additional analyses can be performed on the samples after the DESI-based testing.
DESI uses a stream of ionized solvent droplets, shot at an angle toward the sample of interest. These droplets extract ions from the sample, which are then directed into the mass spec for analysis. Because samples can be analyzed directly without the need for liquid chromatography, DESI is a simpler and faster process than LC-MS. It is also somewhat simpler than MALDI mass spec, which is also commonly used for imaging applications, due to the fact that, unlike MALDI, DESI does not require the addition of a matrix to the sample.
DESI was developed as a commercial mass spec technology by Prosolia. Last year, Waters purchased that company's DESI assets.
Sulibark is a collaborator of University of Texas researcher Livia Eberlin, who was senior author on the PNAS study. A major focus of Eberlin's work has been developing direct ionization mass spec approaches for identifying cancer tissue with the ultimate aim of implementing it as an approach for assessing cancer margins during surgery. Her lab's MasSpec Pen, which Sulibark has also worked on developing, uses drops of water to extract lipids and metabolites from tissue that are then analyzed via mass spec to generate molecular profiles that distinguish between different tissue types.
Sulibark said that he and his colleagues were currently looking into commercialization opportunities for the technology. While imaging mass spec is not yet a clinical technology, he said he believed that as instruments become more streamlined and miniaturized, imaging mass spec will carve out a space as a complement to traditional pathology.
"We really have not changed in our cytologic or pathologic diagnosis very much for 100 years, and the use of molecular analysis of tissues coupled with a nice data-driven approach to modeling molecular pathologic behavior is really going to be the future," he said. "Gene testing is already here and now, but what we would like to do is develop more of a real-time capacity for testing because then we can integrate that information immediately into the patient care plan."
He added that because of the absence of LC and the limited amount of sample prep required, the assay can generate results in a matter of minutes.
The researchers are now collecting additional thyroid FNA from multiple centers that they will use to further develop and validate the assay. Sulibark said they are aiming to collect around 700 to 1,000 samples.