AI-based risk assessment can be effective for personalized blood-based multicancer screening
Around 20 years have witnessed transformative advances in molecularly targeted and immunological treatments for advanced cancer, offering many patients with prolonged survival and quality of life. However, the main factors of cure across diverse cancers remain the stage of diagnosis. The recent progression of blood-based multicancer detection (MCD) assays, together with advances in imaging and artificial intelligence (AI) algorithms, can detect cancer earlier. However, these innovations have health and financial risks, and their increasing availability raises both opportunities and obstacles, which are evident as clinics dedicated to early cancer detection are launched.
A study describes a scenario of a 55-year-old woman who, despite following the latest recommendations for cancer screening and having unknown cancer genetic susceptibility syndrome, decides to pay for a multicancer detection test (Galleri) due to concerns about her risk of cancer. As per the test, a “cancer signal was detected” with the ovary as a top predicted tissue of origin, but subsequent clinical work-up, including high-resolution imaging and the ovarian cancer antigen 125 (CA-125) blood marker, was negative. The text raises questions about how to counsel the individual who has a positive cancer signal on a blood test but appears healthy. It also raises questions on how common such scenarios are likely to be as multicancer detection screening becomes increasingly available.
The Galleri test has shown some effectiveness in detecting early-stage cancers, but also has a high rate of false-positives which can lead to unnecessary invasive tests and anxiety. The ongoing NHS trial aims to determine if adding Galleri testing to standard cancer screening protocols can lead to earlier cancer diagnosis, but it may not be able to account for lead-time bias.
The implementation of MCD screening tests raises the question of whether it is better to apply them to all persons above a certain age or to use AI to develop more individualized risk-based screening strategies. The cancer risk increases with age, and the annual incidence is higher in older people. The positive predictive value (PPV) of a screening test depends on its specificity and sensitivity, as well as the cancer prevalence in the tested population. A hypothetical test with 99% sensitivity at 99% specificity will produce more false-positive results than true positives in a population with only a 1% cancer prevalence. However, if the cancer prevalence increases to 5%, the PPV for the same test will be higher, meaning fewer false positives.
The use of AI-driven approaches can help in integrating traditional and new factors to assess cancer risk. This approach can be helpful in predicting an individual’s future risk of breast or lung cancer by analyzing radiology images of noncancerous tissue. This individualized cancer risk assessment can lead to more effective targeting of blood-based MCD screening to high-risk populations, which can improve PPV.
It is important to discuss the limitations of age-based screening for cancer, which can lead to a large number of false positives and high costs. It suggests that using AI-based risk calculators to identify high-risk individuals could improve the accuracy of screening. Additionally, the use of multicancer detection (MCD) testing for uncertain radiographic lesions could be a useful clinical application with high positive predictive value (PPV).
The use of MCD screening in clinical care presents a significant opportunity and challenge. By objectively assessing personalized cancer risk, it can improve cancer screening strategies, increase predictive power, and reduce unnecessary anxiety and medical workups.
This news is a creative derivative product from articles published in famous peer-reviewed journals and Govt reports:
References:
1. Klein EA, Richards D, Cohn A, Tummala M, Lapham R, Cosgrove D, Chung G, Clement J, Gao J, Hunkapiller N, Jamshidi A. Clinical validation of a targeted methylation-based multi-cancer early detection test using an independent validation set. Annals of Oncology. 2021 Sep 1;32(9):1167-77.
2. Cristiano S, Leal A, Phallen J, Fiksel J, Adleff V, Bruhm DC, Jensen SØ, Medina JE, Hruban C, White JR, Palsgrove DN. Genome-wide cell-free DNA fragmentation in patients with cancer. Nature. 2019 Jun 20;570(7761):385-9.
3. Lennon AM, Buchanan AH, Kinde I, Warren A, Honushefsky A, Cohain AT, Ledbetter DH, Sanfilippo F, Sheridan K, Rosica D, Adonizio CS. Feasibility of blood testing combined with PET-CT to screen for cancer and guide intervention. Science. 2020 Jul 3;369(6499):eabb9601.
4. Placido D, Yuan B, Hjaltelin JX, Zheng C, Haue AD, Chmura PJ, Yuan C, Kim J, Umeton R, Antell G, Chowdhury A. A deep learning algorithm to predict risk of pancreatic cancer from disease trajectories. Nature Medicine. 2023 May 8:1-0.