Integrated Diagnostics has published a study this week in Science Translational Medicine presenting the discovery and initial validation research underpinning its forthcoming proteomic lung cancer diagnostic.
Scheduled for launch sometime this year, the InDi test is intended to distinguish between benign and cancerous lung nodules discovered by computed tomography imaging. In the STM study, the company's researchers and their collaborators demonstrated that the test could identify cancerous nodules with a negative predictive value of 90 percent.
The paper presents data from two separate studies. In the first – a three-site discovery study looking at plasma samples from 143 patients with matched stage IA lung cancer and benign nodules – the researchers built multiple-reaction monitoring mass spec assays to 371 proteins identified as potential markers via analysis of resected lung tumors and literature searches. Running these assays against the 143 patient samples, they developed a 13-protein classifier that distinguished between benign and malignant lesions with NPV of 90 percent.
The researchers followed this with a four-site validation study in which they tested this panel against 52 cancer and 52 benign samples, again achieving an NPV of 90 percent. This study included samples from a new clinical site, Vanderbilt University, on which the panel performed with NPV of 94 percent, indicating, the authors noted, that the panel "was not overfit to the discovery sites."
Since then, InDi and its development partners, which include Caprion Proteomics and the Institute for Systems Biology, have completed additional validation studies testing the panel against nearly 500 patient samples, InDi CEO Albert Luderer told ProteoMonitor. Both these studies and the work presented this week in the STM paper used retrospective samples taken from archives of prospectively collected patients. These patients, Luderer said, were all given full medical follow-ups, with their lesions all histopathologically determined to be either benign or malignant. All cancerous nodules were stage I, he added.
"This was a very difficult trial to put together, and part of the strength of the trial is not only the qualification of the patients, but also the fact that it was multi-center," Luderer said. "In a lot of clinical trialing you see great performance in a single center, but then you can't replicate it when you go to other centers. So we decided very early to force the question of biological variability, center-to-center variability, all those sorts of things, to determine whether or not our signals were strong enough to move ahead."
He noted that while the company plans to take the test to market this year based on this retrospective sample data, it is currently involved in two large prospective validation trials, as well. InDi hopes to report initial data from these trials in late 2014, he said.
While the STM study describes results from a 13-protein panel, Luderer said that the commercial version of the test would use only five protein classifiers, along with six additional proteins included as "normalizers" to help the company identify and eliminate biological variation unrelated to a patient's lung cancer status.
"There is so much individual variability, and by approaching this using normalizers, we are able to reduce that type of uninteresting biological variability to a very low level that allows the disease signal to stand out," he said.
Interestingly, the signal detected by the InDi panel appears unrelated to traditional clinical risk factors such as nodule size, age, and smoking history. While the STM authors present this as a potential limitation, noting that a classifier that integrated all of these factors "would be ideal," Luderer suggested that this could, in fact, enhance the test's value by offering an orthogonal measure to help doctors make treatment decisions.
"We have talked to a tremendous number of physicians, and there is a belief among pulmonologists that there is a tremendous need for additional information to help them triage [patients]," he said, noting that the invasive procedures currently used for determining lung cancer status, like fine needle aspiration and thoracotomies, carry not insignificant risks.
The InDi test, with its focus on NPV, is designed to help physicians more confidently rule out lung cancer in patients and put them on a regimen of watchful waiting, Luderer said.
Beyond its significance to InDi and potential significance to the lung cancer space, the InDi panel – assuming it launches on schedule – also represents something of a landmark for proteomics, as it will be the first proteomic test brought to market on an LC-MS/MS platform.
A number of companies including Vermillion, Aviir, and Crescendo Bioscience have brought immunoassay-based proteomic tests to market, while others like Biodesix, Bruker, and BioMérieux have moved MALDI mass spec-based tests into the clinic. InDi would be the first, however, to launch a multi-protein panel using targeted proteomics on an LC-MS/MS platform.
The company performed the work detailed in the STM paper on an AB Sciex 5500 QTRAP instrument but has since switched to an Agilent 6490 triple quadrupole for the test's commercial launch. Agilent and InDi announced last year a strategic partnership to develop mass spec workflows for protein biomarker diagnostics. As part of that alliance, InDi purchased two Agilent 1290 Infinity UHPLC systems, two 6490 iFunnel triple quads, and two 1260 LCs with ultraviolet detectors (PM 1/13/2012).
In working to move LC-MS/MS-based proteomics into the clinic, researchers have struggled with a number of technical issues, prominent among them assay reproducibility and throughput. According to Luderer, InDi, which plans to offer the lung cancer test out of its CLIA lab, has managed to achieve single-digit coefficients of variation for each of the peptides comprising the panel.
In terms of throughput, the company can currently process several hundred patient samples per week, Luderer said, adding that if demand makes it necessary, this capacity can be expanded significantly.
"This is a new technology, and it is not particularly easy to do the chemistry, so we are moving very conservatively," he said. "Over time, and if the market says we ought to, we can really improve on the throughput."