NEW YORK – Researchers led by a team at McGill University have developed a sequencing-based assay for the early detection of gynecologic cancers, which they are currently validating in a prospective study.
Called DOvEEgene (Detecting Ovarian and Endometrial cancers early), the assay involves targeted sequencing of a brush sample of the uterine lining, coupled with a machine-learning algorithm trained to distinguish endometrial, fallopian tube, and ovarian cancers from both healthy controls and benign disease that might present with the same symptoms.
Ovarian cancer in particular has long been a target for the development of molecular screening tools, because most tumors are detected at late stages with no hope for a cure, and, often, a poor chance for disease control.
"In many healthcare systems, even if you have symptoms, you will go to your family doctor to complain about your symptoms, then you're sent to a specialist, then you are sent to a hospital, and you can see how many steps and hoops you have to jump through until you get a test done," said Ioannis Ragoussis, a codeveloper of the test and head of genome sciences at McGill.
Ragoussis discussed the test and its ongoing clinical validation by the university during a workshop at the annual meeting of the Association for Molecular Pathology last month. The session was sponsored by Agilent Technologies, which he said has been a collaborator.
"The aim is to see whether it's possible to introduce a test that can be a screening test for the entire population and can directly give a result without the patient having to wait for symptoms, and go through all the subsequent [hoops]," Ragoussis said.
For the study, led by McGill oncologist Lucy Gilbert, investigators have recruited several thousand women meeting certain eligibility criteria — age 50 or over with a uterus — regardless of whether they have any symptoms. All participants receive the DOvEEgene test alongside a normal pap smear, saliva sample, blood sample, and pelvic ultrasound, with follow-up at several subsequent time points.
If the test returns a positive result, patients receive additional diagnostic testing. Those with a negative result are rescreened either six months or a year later.
The assay uses duplex sequencing with a high-sensitivity version of Agilent's SureSelect target enrichment reagents to create a panel that covers 32 genes and includes both germline and somatic mutations. Added low-pass whole-genome sequencing allows for the analysis of copy number alteration, and paired sequencing of saliva helps to confirm that mutations detected in the brush sample are truly somatic.
The test algorithm then combines different parameters, like mutation pathogenicity, variant allele frequency, and age to call a case as either cancer or non-cancer.
According to Ragoussis, the goal was to optimize the ability to detect variants at frequencies below 0.1 percent while making the technology transferable to as many labs as possible for population-based testing.
Early analytical testing showed that the assay can detect mutations in its target genes at high sensitivity, with the most alterations occurring in endometrial cancer, fewer in ovarian cancer, and even fewer in control samples from patients with potentially confounding benign conditions.
Training the algorithm against these other diseases, rather than just a general set of asymptomatic controls, was crucial, Ragoussis said, because somatic mutations are known to accumulate with age, even in benign disorders.
He presented validation data at the conference from an initial training and test set, in which the initial version of DOvEEgene detected over 92 percent of endometrial cancers. Ovarian cancer sensitivity was significantly lower, at about 40 percent.
At this point in the ongoing prospective validation study, investigators have analyzed over 4,000 samples, Ragoussis said, with mutational patterns in detected cancers versus controls looking much like they did in the initial analytical validation.
So far, the assay has detected endometrial cancer in 15 individuals, all at stage I or II. "These women have now been treated successfully, even before there was a substantial tumor present," Ragoussis said.
The rest of the cohort is being followed until 2025, when the team will draw its final conclusions.
Ragoussis said that he and his colleagues have not finalized where they plan to set the cutoff for the test in terms of balancing sensitivity and specificity. But for the trial, the assay was locked in at 90 percent specificity. "When the whole study results are evaluated, we will probably readjust, because in this setting, improved sensitivity may be worth lower specificity," he said.
Although sequencing is relatively expensive compared to protein blood tests that are currently available, Ragoussis argued that there is good reason to believe that the health economics will balance out in its favor, considering the high costs of treating advanced gynecologic malignancies. A researcher is working on modeling that right now, he added.
Because of the unmet need for sensitive tools to screen for gynecological cancers, the McGill team is not alone in seeking a molecular test. Newer approaches also gunning for commercial and clinical success include a blood-based assay using extracellular vesicles, pursued by Mercy Bioanalytics.
Companies advancing cell-free DNA methylation and fragmentation patterns for early cancer diagnosis, like Delfi Diagnostics, have also taken some preliminary steps to test these technologies in ovarian cancer.
For all of these, prospective data — like that being collected in the DOvEEgene trial — will be crucial in making the case for clinical utility.