To improve early detection, cancer screening studies need to evolve and integrate multimodal data in the identification of high-risk individuals.
There has been steep increase in understanding how risk factors such as age, gender, genetics, cultural determinants and environmental exposure are associated with different types of cancer. In parallel, technologies that enable cost-effective molecular profiling for cancer detection are evolving quickly. Yet screening recommendations for early cancer detection remain limited to a handful of cancer types, with age and family history of cancer still being the main factors used to determine cancer risk. In the age of precision medicine, when cancer therapies are tailored to the molecular profile of the patient’s tumors, cancer screening must also become more precise and tailored to an individual’s cancer risk.
In contrast to existing single cancer detection and screening tests, such as mammography for breast cancer or human papillomavirus DNA tests for cervical cancer, which have been around for decades, multi-cancer early detection (MCED) liquid biopsy tests represent a new development that can cast a wide net and facilitate the detection of several cancer types simultaneously. Unpublished results of the prospective MCED screening study PATHFINDER2, which enrolled 36,000 healthy individuals over the age of 50, showed that nearly 62% of individuals who had an initial test positive result were confirmed to have cancer. Aggressive cancer types such as pancreatic adenocarcinoma (PDAC) and ovarian cancer were detected at early stages, when patients are more likely to benefit from treatment. However, the test caught only 40.4% of all cancer cases. The relatively low yield of early cancer detection and the high number of cancers not detected by the test are at odds with the need to maximize benefit while minimizing harm through unnecessary testing and diagnostic workup (or lack thereof entirely), which can be met only if high-risk individuals are more accurately identified before screening.
For prostate cancer, current recommendations include optional screening of plasma levels of prostate-specific antigen (PSA) in 55- to 69-year-old men. PSA screening reduces prostate cancer mortality by 13% compared with no screening1. However, of the 16% of PSA tests that showed increased PSA levels requiring further diagnostic efforts, only 24% confirmed a diagnosis of prostate cancer, meaning that a high number of patients, while eligible for screening based on their age, were exposed to unnecessary and costly follow-up screening.
Lung cancer screening has been recommended to individuals by smoking history and age (based on the US Preventive Services Task Force (USPSTF) recommendations for screening in 2013), but more precise methods to identify individuals at risk are being implemented. The identification of screening-eligible individuals using multivariable risk predictive models such as the Liverpool Lung Project-v2 (LLPv2) and the model derived from the Prostate, Lung, Colorectal and Ovarian study (PLCOm2012), which integrate demographic and clinical variables to identify at-risk populations, can improve lung cancer detection compared with the USPSTF recommendations2. Encouragingly, these models are being implemented in national lung cancer screening programs in Canada and the UK2.
Prospective clinical research into multimodal risk-adaptable screening is ongoing for other cancer types. In breast cancer, the WISDOM study in the USA, as well as the MyPEBS study in Europe, compare personalized risk assessment and screening recommendations with the standard-of-care national screening recommendation in the respective country (such as annual mammography in 40- to 74-year-old women the USA). Personalized risk is determined using a multifactorial risk score (age, breast cancer family history, race/ethnicity, breast density, body mass index and menopausal status) and saliva-based genetic testing, based on which an individual screening schedule is created. These are examples that likely advance cancer screening and detection methods through a focus on patients’ needs, implementing available, state-of-the-art tools for personalized screening.
Artificial intelligence (AI)-powered extraction and analysis of data from medical or electronic health records (EHRs) (also referred to as ‘AI biopsy’) can predict an individual’s risk of several diseases3 — a potentially cost-effective and accessible approach for population-based screening. Although this approach can be effective in predicting the risk of even one of the most aggressive and hard-to-detect cancer types, PDAC its application depends on the quality and density of EHRs4 and requires availability of an appropriate clinical infrastructure. Still, the integration of AI biopsies in multimodal cancer risk screening will be crucial to leverage their cost-effective potential in identifying patients for screening.
Emerging research into the human exposome has surfaced complex interactions among host biology, cancer risk and environmental and lifestyle exposures5 that could provide clues for more accurate risk assessment of cancer types, including early-onset cancer. Research into precancer biology will offer further opportunities to determine cancer risk in healthy individuals. As such, research into organ-specific aging shows that blood-based DNA methylation analysis can be used to predict organ-specific disease risk6. Microenvironmental drivers of precancer initiation, including very early anti-tumor immune responses, could potentially be measured years before cancer onset and inform screening frequency and prevention strategy.
Cancer therapy has already shown that a one-size-fits-all approach rarely works for everyone. Research now shows that multimodal risk identification beyond age can lead to more efficient screening for early cancer detection than relying on the current minimal standard. A minimized risk for overdiagnosis and overtreatment will make the swift implementation of effective screening strategies in routine care more likely, as seen in lung cancer. When designing clinical trials and implementation studies to validate new methods for early cancer detection, the opportunity for a precision screening approach in at-risk individuals must not be missed.
