Johnson ATC, et al. Abstract 9004. Presented at: ASCO Annual Meeting (virtual meeting); June 4-8, 2021.
‘Electronic nose’ accurately detects ovarian, pancreatic cancers
Johnson ATC, et al. Abstract 9004. Presented at: ASCO Annual Meeting (virtual meeting); June 4-8, 2021.
An electronic, odor-based tool accurately distinguished pancreatic and ovarian cancer specimens from benign disease and control specimens by analyzing vapors emanating from blood samples, according to results of a small, preliminary study.
The results, presented during the virtual ASCO Annual Meeting, indicate the tool could represent a noninvasive approach to screen for hard-to-detect cancer types, researchers noted.
“We have been working on the issue of early detection by liquid biopsy for quite some time, and although we have made inroads with current approaches, they have not been sufficiently sensitive,” Erica L. Carpenter, PhD, director of the Liquid Biopsy Laboratory and Core and research assistant professor at Perelman School of Medicine at University of Pennsylvania, said during an interview with Healio. “This study was an interesting opportunity because it is a whole new way of trying to detect an early tumor, and the hope is that it will yield additional information.”
Carpenter and colleagues assessed the ability of the electronic-nose tool to distinguish vapor characteristics of plasma samples from 93 individuals, including 20 women with ovarian cancer, 20 women with benign ovarian tumors and 20 age-matched, cancer-free women, in addition to 13 patients with pancreatic cancer, 10 patients with benign pancreatic disease, and 10 age- and sex-matched controls.
According to study results, the tool accurately identified ovarian cancer with 95% accuracy and pancreatic cancer with 90% accuracy, including all eight patients with early-stage disease.
Healio spoke with Carpenter and fellow University of Pennsylvania researchers Jody R. Piltz-Seymour, MD, and Christopher Kehayias, PhD, about the study, what they found, what surprised them most about the findings and what future research will entail.
Healio: What prompted this research?
Piltz-Seymour: First, there is a huge unmet need for screening tests for so many cancers. For ovarian cancer specifically, we know about 20,000 women will be diagnosed this year in the U.S. and about 14,000 will die of the disease. If caught early, these women have an excellent prognosis, with 5-year survival rates of 98% for local disease and almost 90% for regional disease. Unfortunately, it is not usually caught early because there is no current strategy to screen for ovarian cancer. When caught with distant metastases, the survival rate is only 31%. This is after harsh treatment that these women go through. The same can be said of pancreatic cancer, but the situation is even worse because pancreatic cancer is almost universally caught late with a poor prognosis.
Second, I have watched multiple generations in my family succumb to both ovarian and pancreatic cancers. Despite their diagnoses being separated by decades, no significant advances had been made that may have identified their cancers earlier and at a more treatable stage. I am hoping that the next generation will have a different story to tell.
Healio: Can you briefly describe how the tool works?
Kehayias: When cells divide, some volatile organic compounds (VOCs) are generated as byproducts and then absorbed into the blood stream. Tumor cell metabolism is different from that of healthy cells, hence the reason why VOC contents of plasma from patients with cancer vs. healthy individuals are different. Our sensor technology takes advantage of this principle by differentiating between blood-derived odor signatures associated with malignant forms of cancer, benign cancer and no cancer.
The electronic nose sensor design draws inspiration from mammalian olfaction architectures, which contain many different types of olfactory receptor neurons that each respond differently toward a given odor molecule, collectively producing an olfaction “fingerprint” unique to that molecule. Our sensors are based on carbon nanotubes with extremely sensitive electronic properties combined with single-stranded DNA to induce chemical affinity for VOCs. The ensuing interactions between VOCs and the DNA-nanotube structures evoke an associated electrical response from the sensors that are highly dependent on the particular base sequence of the DNA. The sensor arrays are equipped with ten different groups of sensors, each functionalized with a different DNA oligomer to enable chemically diverse responsivity toward target molecules, analogous to the multiple different types of olfactory receptor neurons found in our noses. A gas delivery system is used to sequentially expose our sensors to VOCs from plasma samples. The resulting electrical output is processed using pattern recognition tools that enable classification of samples according to malignant, benign or healthy diagnostic groups.
Healio: What are the key takeaways from the study?
Carpenter: This noninvasive test that is based on a small amount of a patient’s blood was able to distinguish not only between blood that had been obtained from a healthy individual vs. an individual with cancer, but also between cancers and benign conditions both in the ovarian and pancreatic settings. The inclusion of noncancer disease controls is really important in a study like this because patients who have, for example, pancreatitis, pancreatic cysts or other diseases that are not cancer often have a lot of inflammation and other comorbidities that can confound our ability to accurately detect cancer. One of the things that we do not want with a test like this is to have a false-positive result.
Healio: Did any of the findings surprise you?
Carpenter: One reason I was so surprised and impressed with how well the test performed was that some of these noncancer disease conditions occur in patients with type 1 diabetes who are at a higher risk for developing a tumor. Those are the patients we screen more often because of inflammation, illness and other confounding factors, but they actually can be the most difficult to screen. I am really excited now to move forward and continue to build the size of our cohorts. Hopefully we will continue to be impressed and surprised as we extend to other cancer types.
Piltz-Seymour: This is all very early data that need to be verified in larger studies, but when the physicist showed me the data, I had never seen curves for screening with such little overlap between diagnostic groups. It was breathtaking to see. There is hope that we are on the right track. We are taking a completely different strategy by harnessing the power of what may be in the VOCs in the headspace air above our blood samples and using that as a substrate to test for differences in the diagnostic groups.
In one sense I feel like I should not have been surprised, because my original interest in this project stemmed from work done in Sweden by a researcher using dogs to distinguish plasma samples of patients with ovarian cancer from those with benign ovarian disease and controls. He was able to show sensitivities over 95%, and that was the original stimulus for looking further into using VOCs as a substrate.
Healio: What new information do the data provide for our oncologist readers?
Piltz-Seymour: If reproduced, the data have shown that this technology may be used to achieve our goals of early detection. We have some very preliminary work that indicates it may be useful in other cancers, as well. It is a technique that may be widely applicable.
Healio: Did you encounter any challenges with the research?
Carpenter: This type of research requires multidisciplinary coordination and collaboration. So far, it has gone really well and we are only just getting started together. However, it can be difficult to do. Not that many years ago, breaking down the silos between different divisions could be just as challenging as doing the work itself. We are all from different groups here within University of Pennsylvania, so it is a challenge to get together and concisely describe and analyze the data, but at the end of the day, this is living proof that it can be done and this is the way to conduct research like this.
Kehayias: The chemical-sensing applications that we were looking into prior to measuring blood samples were based mainly on VOCs with known properties and predictable interactivity with our electronic nose sensors, allowing us to design our experimental procedures accordingly. Blood plasma, on the other hand, generates a mixture of many different types of volatiles, which made optimizing experimental parameters challenging. We eventually developed effective techniques for measuring plasma samples that were partially inspired by methods used for gas chromatography-mass spectrometry.
Piltz-Seymour: Another challenge is funding. It is a pity that with research, so much time, effort and energy must be spent on finding funding sources. At the beginning, it was a real challenge to gain traction and credibility in the work that we were doing because it was a little bit “outside the box” and people were not very interested in it.
Healio: Are there plans for additional research?
Carpenter: Yes. We are expanding upon the ovarian and pancreatic cancer cohorts that we reported on at ASCO to see if we can replicate the results. We recently applied this technology in the setting of prostate cancer and those results are very promising so far. That is a third solid tumor setting in which we are seeking to expand the number of samples, including from men with pathologically confirmed prostate cancer, men with non-cancer prostate disease and healthy controls. We also have a large sample set of thousands of patient samples in my laboratory, and we are working together to find the best time, resources and funding to start to expand to other solid tumors.
One of the problems that we need to solve involves identifying the type of cancer the patient has. That may be a logical extension of this work — to work with the artificial intelligence team to reassess the data we have and ask questions such as: Can the model not only distinguish between cancer and noncancer, but also determine where the patient’s primary tumor is?
Piltz-Seymour: We certainly want to evaluate how early in the process we can detect cancers and be able to see if we can distinguish between early vs. late-stage and local vs. metastatic disease.
Kehayias: We would like to begin clinical trials using our sensor technology and are actively making upgrades to our hardware to accommodate the ensuing needs. Our current system is large, occupying a tabletop roughly 1 meter by 2 meters. We are working with our partners at VOC Health to innovate and commercialize a miniaturized system that can fit entirely within a briefcase. This new portable system will be able to measure samples more rapidly and with greater ease, thus making it ideal for clinical studies. We also have increased the number of DNA-carbon nanotube sensor groups supported by our electronic nose arrays to enhance their overall discrimination power.
American Cancer Society. Key statistics for ovarian cancer. Available at: https://www.cancer.org/cancer/ovarian-cancer/about/key-statistics.html. Accessed June 17, 2021.
American Cancer Society. Survival rates for ovarian cancer. Available at: https://www.cancer.org/cancer/ovarian-cancer/detection-diagnosis-staging/survival-rates.html. Accessed June 17, 2021.
Johnson ATC, et al. Abstract 9004. Presented at: Presented at: ASCO Annual Meeting (virtual meeting); June 4-8, 2021.
For more information:
Erica L. Carpenter, PhD, can be reached at email@example.com.
Christopher Kehayias, PhD, can be reached at firstname.lastname@example.org.
Jody R. Piltz-Seymour, MD, can be reached at email@example.com.