We report early real-world experience with a validated MCED test designed for cancer screening and capable of clinical implementation. Although the MCED test detects cancer across all stages, it provides early detection (stage I). All three were asymptomatic, so early diagnosis was unlikely, and two of them had no risk factors other than age. In all three cases, the CSO was proven correct by pathology and served as a guide for efficient diagnostic evaluation. The time to diagnostic recovery ranged from 28 to 68 days, consistent with that reported in the PATHFINDER study.12. All three were eligible for and received curative treatment with guideline-compliant care.
These people's experiences highlight the potential of this technology to detect early cancer in asymptomatic people and demonstrate the ability of CSO predictive capabilities to efficiently achieve diagnosis. Regarding the specific cancers discussed here, ovarian cancer tends to develop in the late stages.16difficult to diagnose because symptoms are nonspecific or absent17screening has primarily been considered in a subset of genetically high-risk individuals, and there are notable limitations in the performance of screening instruments (e.g., transvaginal ultrasound).18, 19, 20. Similarly, early-stage RCC is usually asymptomatic and often detected incidentally on imaging tests, making the potential for overdiagnosis significant.twenty one. In the case described here, the person has a more aggressive histology and more aggressive behavior is predicted. Currently, there are no recommendations regarding her RCC screening for individuals with average risk. Finally, there are no routine screening programs or tests for oropharyngeal cancer that go beyond what can be seen during a routine oral exam at the dentist or self-examination.15. HPV is a risk factor for oropharyngeal cancer, but in contrast to cervical cancer, there is no approved HPV screening test for pharyngeal cancer.15. Furthermore, given that single cancer screening tests for less common cancers may not be feasible, detection of less common cancers such as oropharyngeal cancer in the real world is Particularly noteworthy.
If screening does not occur, and early cancer is not detected, it may progress to a more advanced stage before clinical symptoms leading to diagnosis appear, at which point the prognosis may be less favorable. The stage dependence of survival outcomes for these cancer types suggests that these three cases are likely to have good long-term outcomes (local stage of these cancers; Survival outcomes at distant stages were ovarian (93.1%, 74.2%), 30.8%, RCC, 93.3%, 74.7%, 15.7%, and oropharyngeal SCC, 83.1%, 77.8%, 48.7%), respectively.22,23.
The technology underlying MCED testing relies on the detection of circulating cfDNA associated with tumors. Therefore, not all tumors and tumor types release cfDNA in amounts above the clinical limit of detection (LOD), and therefore not all cancers can be detected with this technology. For example, in the CCGA study, the overall sensitivity for head and neck cancer was 85.7%, for ovarian cancer it was 83.1%, but for kidney cancer it was only 18.2%. Kidney cancer is known to be one of the less excretory tumor types.Detection rate is less than 4% for low-grade prostate cancer associated with lazy behavior7. Her three cases described here benefited from the fact that her tumors released her cfDNA at levels that exceeded the clinical LOD of this test. Some evidence suggests that such tumors are associated with the potential for aggressive behavior even at early stages.24,25.
Several considerations should be kept in mind when weighing the clinical insights supported by these cases. First, in these three cases, the CSO calls corresponded to the type of tumor diagnosed. During comprehensive clinical evaluation and at least 1 year of follow-up in all 3 cases, there was no evidence of a second type of cancer, indicating that the tumor shedding was due to the diagnosed and treated cancer. Masu. Additionally, while the individual cases presented here are intended to serve as case descriptions, they nevertheless represent only a small subset of the larger set of individuals who underwent this test. It is appropriate to recognize that.
MCED represents a new paradigm with the potential to address important unmet needs in cancer screening. By combining next-generation genome sequencing and machine learning, MCED tests can detect multiple cancer types, including cancers that are not widespread enough to allow efficient single-cancer screening.26,27. Because this test detects cancer signals common to multiple cancer types, individual cancer prevalence across multiple cancers can be aggregated to improve screening efficiency, resulting in currently approved It has a much higher PPV and overall cancer detection rate than conventional screening tests.2,28. Additionally, machine learning algorithms continuously learn from new data of the type presented here, allowing them to continually improve the performance characteristics of their tests.
Machine learning is a subcategory of the broader field of artificial intelligence that uses algorithms to automatically learn insights, recognize patterns in data, and apply that learning to make increasingly better decisions. To do.29. In this case, his more than 15,000 individuals in his CCGA study, which enrolled participants between 2016 and 2018, to learn which cfDNA fragments may originate from cancer cells. A classification algorithm was first trained based on sequence data from .6,7. The study included 6,670 people without cancer and 8,584 people with cancer, and also recorded cancer type and comorbidities. The first step in the classifier training phase was to determine the correct way to encode DNA methylation status into a computer-readable (“representation”). Second, the algorithm compared the methylation patterns of cancer-free individuals and individuals known to have cancer in CCGA to derive common cancer signals (“learning”). This cancer sign is rarely observed in people who are not known to have cancer. Finally, the algorithm assigned each individual a score estimating the probability of having cancer, and then assigned each of these probabilities to one of two bins: a cancer signal was detected, i.e. Test positive, or not detected, i.e. test negative (“Thresholding and Scoring”). Once the classifier is trained in this way and passes through the representation, learning, and scoring stages, it is tested and validated against additional data it has not yet seen. If the classifier returns a positive test, a second algorithm is triggered to learn which cells the cancerous cfDNA fragments originate from and predict the CSO. The training stage runs on 1600 computer processors and takes 4 hours. Daily prediction, on the other hand, runs on 48 processors and takes 1 minute. This approach was chosen because it allows for a continuous learning environment and allows the classifier to be trained on more diverse data, improving performance over time.
Unlike current single-cancer tests, which are calibrated to maximize sensitivity and therefore have high false-positive rates, MCED tests aim for high specificity and very low false-positive rates (<1%). Designed and promised to minimize potential harm. Importantly, the MCED test used in these cases can predict the origin of cancer signals and facilitate streamlined diagnostic evaluation. Although these three cases do not stand alone as evidence for clinical use, they demonstrate the power and potential of the test for early diagnosis and how new AI-based technologies can be directly applied to real-world clinical settings. We provide an example of how it can be done. Optimize patient care. These cases should be reviewed based on the robust clinical trial data and ongoing accumulation of real-world evidence supporting its clinical use as an LDT. When used at a population level, MCED testing has the potential to reduce cancer mortality by stopping cancer at an early stage.28.
