Multicancer detection tests: What we know and what we don’t know

Abstract

The concept of blood-based multicancer early detection (MCED) tests has generated much excitement, in part because of the potential of such tests to reduce cancer mortality by encompassing cancers for which screening is currently not available. A review in this issue of CA: A Cancer Journal for Clinicians, largely authored by members in the Division of Cancer Prevention at the National Cancer Institute (NCI), addresses the current status of the field.¹ The authors convey a reluctance to refer to the field as MCED. In their view and that of others, the evidence to date does not support substantial performance in detecting cancer at an early stage.² Therefore, instead, they use the designation multicancer detection (MCD) tests. The authors describe a strategy for MCD tests adopted by developers, consisting of first detecting a cancer signal based on shared biomarkers across cancer types, followed by assessment of the tissue of origin based on another set of biomarkers. The review includes a list of developers of MCD tests and the performance of tests for which data have become publicly available based on their positive and negative predictive values. The authors also provide details of the NCI Vanguard program aimed, in the short term, at testing the performance of MCD platforms they have selected among applicants and, in the longer term, at conducting prospective, randomized clinical studies.

Although the review provides an assessment of the current status of the MCD/MCED field, there is much that we do not know and that remains to be determined. From an effectiveness point of view, the optimal number of cancer types to be included may be debated. Currently, screening is available in the United States for lung, breast, colon, cervical, and prostate cancers. Screening is also available for gastric cancer in Asian countries, where the incidence is high. Although MCD tests have the potential to encompass a much broader range of cancers, notably including cancers for which screening is not available, it is clear that a relatively small number of cancers account for the vast majority of cancer deaths. American Cancer Society cancer statistics 2024 data for US cancer mortality project that five cancer types account for greater than 50% of cancer deaths.³ For men, they include pancreas and hepatobiliary cancers and, for women, pancreas and ovarian cancers. Given that an MCD test may vary in its performance by cancer type in terms of sensitivity and specificity, overall test performance may degrade with attempts to universally cover a vast number of cancer types. Moreover, for common cancers for which screening strategies are recommended, should MCD tests result in improved positive predictive value of screening programs? For other malignancies, the underlying cancer biology or treatment approaches may obviate any benefit of an MCD test. For example, the authors point out that hematologic malignancies comprised 57% of early stage diagnoses in the Pathfinder study (ClinicalTrials.gov identifier NCT04241796)⁴ and that mortality gains are unlikely to come from the diagnoses of these cancer types. It may be argued that, because MCD tests are developed based on a comprehensive search for biomarkers, in addition to studying a set of molecular markers that identify cancer tissue of origin, it would be beneficial for MCD tests to encompass biomarkers that correlate with lethality, such as invasiveness and escape from immune surveillance, which are relevant for prognostication.

As the authors note, the underlying incidence of a cancer influences the positive predictive value of a screening test. If it were possible to identify individuals who are at higher risk for cancer either through their clinical, environmental, behavioral, or social determinants of health characteristics, then MCD testing would be expected to be of greater effectiveness. A recent study applied artificial intelligence methods to clinical data from several million individuals in Denmark and in the United States, resulting in a risk profile that, when applied, would improve the ability to design realistic surveillance programs for individuals at elevated risk.⁵ Populations with higher cancer risk caused by various exposures would be another rich opportunity to build an evidence base for the clinical utility of MCD.

A critical question remains around the performance requirements for MCD tests for their implementation in clinical practice. A recent publication covered the Early Detection Research Network's Phases of Biomarker Development for the rigorous evaluation of novel early detection biomarkers.⁶ Criteria include sufficient sensitivity in a prospective screening setting and a shift in detection to early curable stages, leading to clinically significant mortality benefit. The latter has yet to be demonstrated for MCD tests. Whether it should be a requirement may be debated, given the need for randomized screening trials at substantial cost and, with survival being a trailing metric, requiring long follow-up to ascertain, by which time the technology may have very well moved on. Models to evaluate clinical utility with alternative trial designs are needed. The Firefighters Cancer Registry Act directed the National Institute for Occupational Safety and Health and the Centers for Disease Control and Prevention to administer a cancer registry for firefighters, a population with known higher cancer risk. This registry could function as a database of MCD testing and provide real-world data to inform policy (https://www.cdc.gov/niosh/firefighters/registry/aboutnfr.html).

Because we do not know the clinical utility of MCD tests, at their current level of performance, we also do not understand how these tests should be optimally priced relative to their clinical value. The Pathfinder study required 473 tests to detect one individual with early stage cancer; the ratio for the DETECT-A study (Detecting Cancers Earlier Through Elective Mutation-Based Blood Collection and Testing) was one patient per 1239 tests. The downstream costs of false-positive tests, the quality of life-years gained by earlier detection, the cost reduction through avoidance of expensive interventions for advanced disease, and other economic outcome measures are unknown.

The implications of a negative MCD test and how often MCD tests should be administered are currently undetermined. There is concern that an individual who has a negative MCD test may forego recommended screening, although MCD tests are not considered an alternative to current screening modalities. Patients who have a positive MCD test but for whom further diagnostic testing fails to detect cancer are left with the troubling question of whether the MCD test was a false-positive result and no cancer is present or whether repeated diagnostic testing is needed in the event that occult cancer is indeed present. Moreover, how long is the reassurance of a negative MCD test good for, and how often should it be repeated since it is only a snapshot in time? Is there value in serial measurements wherein the predictive capability lies not in a threshold test value but in a rising pattern? A trajectory for such repeated tests may be beneficial. In a cohort study of pancreatic cancer, levels of CA 19-9 increased exponentially starting at 2 years before diagnosis,⁷ pointing to a potential benefit of establishing trajectories for biomarkers.

The authors largely represent the NCI’s perspective on test development. However, multiple federal agencies have regulatory interest in MCD, including the US Food and Drug Administration and the Centers for Medicare and Medicaid Services, as well as commercial insurance entities, along with patient advocacy organizations and unions in industries with occupational exposure. The adoption of MCD tests into clinical use is a microcosm of the larger discourse on how transformative technologies are both nurtured to maximize benefit and regulated to minimize risk.

MCD tests are representative of emergent technologies arising out of artificial intelligence. They incorporate biologic intelligence through the use of genomic, proteomic, and metabolomic biomarkers and other types of biomarkers. And their adoption in clinical practice will require emotional intelligence as part of shared decision making with consumers when the evidence base concerning the risk and benefit is still in formation. A cautious approach would be to first explore the value of MCD tests for individuals at increased risk for multiple cancer types, such as heavy smokers who have an increased risk not only for lung cancer but also cancers of the throat, esophagus, liver, and colorectum among others. An enhanced coverage policy will be needed that provides for necessary downstream testing and clinical follow-up while evidence of improved outcomes is sought. The potential for MCD tests to save lives through earlier detection of cancer is real, but we cannot be satisfied with a minimal viable product.

Sam M. Hanash reports research support from Roche Diagnostics and Exopert; honoraria from Abbott Diagnostics; consulting fees from GLG and Guidepoint; has filed an intellectual property patent pertaining to cancer markers for multiple cancer types; has a patent pending (Biomarker panel for assessment of lung cancer risk and of indeterminate nodules); and has filed a trademark application (Pan cancer marker panel) outside the submitted work. Peter P. Yu disclosed no conflicts of interest.