NEW YORK – ClinSpec Dx is developing an infrared spectroscopy-based test for diagnosing brain cancer patients at an early stage and aims to make the assay available for clinical use in Europe within the next two years.
The company spun out of the University of Strathclyde earlier this year and in partnership with the University of Edinburgh recently completed a trial of the test involving 400 patients.
ClinSpec Dx's test relies on using infrared spectroscopy to look at the chemical makeup of a person's blood sample, as well as an artificial intelligence tool that can determine the likelihood that the patient has a brain tumor. The company sees an unmet clinical need for an early-stage brain cancer test, especially one that is not dependent on a particular biomarker or set of biomarkers identified from a liquid biopsy.
"Cancers are heterogeneous," said ClinSpec Dx's CTO Matthew Baker. "If you are looking at one single biomarker, then that biomarker is probably only prevalent in a subset of the population that has brain tumors," he said. Given the high number of different kinds of brain tumors — 120, according to the National Brain Tumor Society — it's unlikely that a certain biomarker will be prevalent in all of those diseases. "If a single biomarker is prevalent, it's not going to be very specific," he added.
Baker leads the Spectral Analytics Laboratory at the University of Stratchclyde, while he spends half his time as CTO of ClinSpec Dx. The company currently has five full-time employees and is housed at the university's Technology & Innovation Center in downtown Glasgow. Following a proof-on-concept study in 2014, the company raised £560,000 ($720,000) from Scottish Enterprise and has raised £1.6 million in total to date, Baker said.
Last month, the company detailed its approach in a paper in Nature Communications. The method involves the use of attenuated total reflection (ATR)-Fourier transform infrared (FTIR) spectroscopy to detect brain cancer. The assay was developed together with artificial intelligence tools using data from retrospective patient cohorts. Once the spectra are obtained via the method, the spectral datasets are analyzed using machine learning tools trained to identify brain cancer signals in an unknown population. As noted in the paper, ClinSpec Dx has worked to improve the throughput of the approach by moving detection onto silicon internal reflection element (Si IRE) slides that allow the preparation and analysis of multiple samples.
"We made it high-throughput, batch-based, and we made the workflow fit with the clinic and how they currently work," said Baker of the innovations. "We upped our game a bit."
New Edinburgh study
In February 2018, the firm placed its test, called the ClinSpec Dx Triage ID system, at Western General Hospital in Edinburgh to undergo feasibility testing. Data from the study was discussed at the National Cancer Research Institute Cancer Conference, held in Glasgow earlier this week. Of the 400 tested, 40 were found to have a brain tumor. The investigators were also able to correctly identify 82 percent of brain tumors and to correctly identify 84 percent of people who did not. In the case of glioma, the most common form of brain cancer, the test was accurate in 92 percent of cases, they said.
Paul Brennan, a senior clinical lecturer and neurosurgeon at the University of Edinburgh, gave a talk at the conference about the hospital's experience with the test. He said a paper detailing the findings is forthcoming.
"I had a healthy degree of skepticism at the beginning because there are plenty of people developing blood tests, but what was different about their work was it wasn't dependent on an assumption about the biology;" Brennan noted. "There's a lot of cell-free DNA work and a lot of people are looking for proteins which they think correlate with the biology of diseases," he said. "But the spectroscopy test is assaying the whole serum," he added. "It's different."
Having a test on hand that can tell clinicians which patients are likely to have brain cancer at an earlier stage could quicken their diagnoses, Brennan underscored. Sometimes it takes months to diagnose patients, who may see a general practitioner multiple times before finally getting diagnosed in emergency departments. As symptoms can be vague, patients are often sent home to see if their condition worsens, or referred to a neurologist for further observation, though some are immediately sent to emergency departments to be scanned. These decisions are made based on cost and resources but having a new diagnostic test within that pathway could change the process for who gets scanned, and how quickly they are subsequently treated, said Brennan.
"You need to be able to stratify patients and say this patient might have a brain tumor," said Brennan. "By putting the test in that pathway, it's obvious how it can influence decision-making."
What works well is his opinion, is the combination of knowledge on how to apply such a test, with the data from infrared spectroscopy, and the use of machine-learning tools. Brennan worked together with ClinSpec Dx to refine its AI algorithms. "We trained lots of different algorithms to get this [to be] successful," said Brennan. "The key is what questions you ask and what kinds of groups of patients do you want to compare together."
The next step for his team at the University of Edinburgh is to continue its clinical evaluation using data from 600 more patients. Brennan said they had been collecting data as part of this new effort since the beginning of the summer. The patient group this time is slightly different, however. This time, they are specifically testing people who are referred to the hospital for scans. Oftentimes, this patient group is not diagnosed at this stage and are later diagnosed in the emergency department. The use of the ClinSpec Dx test could change that.
"We are thinking of patients who slipped through the net with their general practitioners," he said. "We are confident the test will perform well because it's specific and sensitive, but we need that evidence for us to take it out to general care."
The ultimate goal is to make the test available to Scottish patients within the next 12 to 18 months, he noted, either as part of routine clinical care or via a research study. Baker also said that the company intends to have its test cleared for clinical use in Europe in 2021. The path to clearance however is currently obscured by new European regulations covering medical devices and in vitro diagnostics that will come into effect over the next three years, as well as questions around how the UK will operate within a European regulatory framework once it leaves the EU, now anticipated by the end of January 2020.
"Our next step is regulatory approval," said Baker. "But we need to know the regulatory landscape."
Baker noted that ClinSpec Dx is a platform technology and is not based on particular biology or conditions. "It could be applicable to lots of diseases where there is a clinical need," he said of the firm's future aspects. "A major purpose of the company is to expand out into cancer detection," he added. Brennan said that he is working together with the firm to develop other tests that might inform clinicians whether or not cancer is present elsewhere in the body.
"These will have a massive impact," he said. "If this gets up and running, it will really transform the way we deliver healthcare."
An outstanding question is whether catching patients with brain cancer early will lead to any benefit in terms of outcome. Brennan noted that going forward the investigators will also track patients diagnosed with the test as to how any early stage treatment impacts their prognosis.
"How that affects outcome is an important question, but one we will be able to address once we've got the intervention," he said. "All we know at the moment is how delayed people are."
Willie Hamilton, a professor of primary care diagnostics at the University of Exeter in the UK, said that he is aware of the team's initial studies though he is not affiliated with ClinSpex Dx. Hamilton said there is a need for a simple test to triage out low-risk-but-not-no-risk patients, largely in primary care. He said that he will continue to follow the work as it progresses.
"We don’t know if it's better than biomarkers," said Hamilton. "If further studies on spectroscopy are successful, then such a test would be welcome," he said.