The Challenge of Proving the Benefits of Multicancer Early Detection Tests: An NEJM Article

We are on the threshold of a major change in cancer screening. Recent years have seen an explosion in the development of multicancer early detection(MCED, Multi-Cancer Early Detection) tests designed to screen for several cancer types simultaneously.

MCED tests measure blood-based biomarkers, including mutations, methylation, and fragmentation of cell-free DNA. So far, the most advanced published studies examining these tests have been large, nonrandomized interventional trials.

*

The following text is a translation of an article published on 20 May 2024 in NEJM, entitled “Multicancer Early Detection Tests — Keeping a High Bar for Evidence of Benefit” (Multicancer Early Detection Tests — Setting a High Standard for Evidence of Benefit).

*

The new era promised by MCED tests is appealing. Instead of scans, endoscopies, and swabs searching for individual cancers, a blood test could be repeated annually. This could simplify screening and improve access. Beyond breast, cervical, colon, and lung cancers already screened for, it could detect dozens of cancers for which no screening strategy currently exists.

In 2021 and 2022, more than 140 thousand participants aged 50-77 in England joined the randomized controlled NHS-Galleri trial to assess an MCED test developed by GRAIL. The test was designed to supplement standard screening, not replace it. The primary objective is “to demonstrate a statistically significant reduction in the incidence rate of stage III and IV cancers diagnosed in the intervention arm compared with the control arm,” with final results expected in 2026. NHS England initially planned to extend testing to 1 million people in 2024 based on the trial’s first-year results, but recently canceled this plan, saying the early results were not sufficiently convincing. The specific metrics assessed were not made public.

This development surprised some cancer screening experts. For decades, the reference standard for evaluating screening interventions has been a randomized controlled trial (RCT) measuring cancer-specific mortality. Many health authorities apply this strict standard when recommending screening because it involves healthy people, most of whom will not benefit and some of whom will be harmed.

Minimum criteria for introducing disease screening, such as those established by Wilson and Jungner, require that:

1. the disease exists in an asymptomatic form;
2. the screening test can identify asymptomatic disease;
3. treating asymptomatic disease improves prognosis; and
4. the test is feasible to implement.

So far, MCED evaluations have progressed only on the second criterion: showing that they can detect asymptomatic cancers.

Commercial interests appear to influence evidence standards for MCED tests. Advocates argue that RCTs must be completed quickly and need not use cancer-specific mortality as the primary endpoint because sufficient outcome data would take too long to generate. Some reasons are sound, including the rapid evolution of the technology. Less persuasive reasons include the expense of RCTs (such investments are generally worthwhile) and people dying in the meantime (there is no evidence that these tests could prevent that outcome). At the other extreme, some researchers call for all-cause mortality as the primary endpoint in MCED trials. This reflects both the expectation that effective MCED testing could reduce all-cause mortality more readily than single-cancer screening, and concern that a positive result, whether true or false, could lead to diagnostic procedures resulting in death.

MCED advocates argue that, if a test reduces the absolute incidence of advanced cancers, it is reasonable to assume it could reduce cancer mortality. Advanced cancer incidence is indeed a more robust endpoint than the proportion of cancers diagnosed at an advanced stage, which can be affected by overdiagnosis and other biases. However, our team’s research shows that the extent to which advanced cancer incidence serves as a surrogate for cancer mortality varies by cancer type and cannot be evaluated for many cancers targeted by MCED tests. One concern, supported by commercial data, is that cell-free-DNA-based MCED tests may preferentially detect early-stage cancers that appear to have already begun spreading. A test could then appear to reduce stage III and IV cancer incidence while many cancers classified as stage I or II because of testing still carry a high risk of death. A possible solution is to measure “updated stage,” counting participants whose initial stage I or II cancers later recur at an advanced stage toward the primary endpoint of stage III or IV cancer.

Besides using a stage-based endpoint, NHS-Galleri departs from standard RCT design because blinding — whether participants know their assigned group — varies by trial arm and test result. Only intervention-arm participants with a positive MCED result will be unblinded. This choice encourages participation in annual blood testing but prevents study of “false reassurance,” a potential drawback. Someone with a negative mammogram may reasonably believe they do not have breast cancer, whereas someone with a negative MCED result may mistakenly believe they have no cancer at all and delay standard screening or seeking care for possible cancer symptoms. To establish MCED’s benefit–harm balance, researchers should communicate results to all participants, mimicking real-world conditions in which both positive and negative results may bring benefits and harms.

The US National Cancer Institute is planning an RCT that can evaluate multiple MCED tests, include a control group, and use the reference-standard endpoint of cancer-specific mortality. The pilot phase will begin in 2024. This may be the only trial sufficiently powered to determine whether an MCED-based approach can reduce cancer mortality. A potential threat is that MCED tests are already marketed in the United States, making it harder to recruit participants willing to accept random assignment and potentially causing contamination through MCED testing in the control group.

When evaluating MCED tests, health authorities must maintain a high standard before concluding that evidence demonstrates benefit. Going forward, I believe MCED tests should not be implemented without evidence of reduced cancer mortality from trials in which commercial entities neither influence final design or conduct decisions nor participate in data management or analysis. Once at least one MCED test has been shown to reduce cancer mortality, carefully designed comparative studies may sometimes replace additional large RCTs when evaluating new tests. Similarly, although effectiveness cannot be studied in every country, the medical infrastructure in a particular geographical setting must be sufficiently robust to manage test results as in successful RCTs.

The cost of currently marketed MCED tests, reaching approximately 1000 dollars per person for some tests, could strain healthcare systems. Companies are lobbying the US Congress for Medicare coverage. Widespread implementation without an RCT showing reduced cancer mortality is concerning, given the hard-learned lessons about careful introduction of screening. Overdiagnosis of prostate and thyroid cancers and debates over changing breast screening guidelines offer instructive examples. Nevertheless, translating recent biotechnology advances into cancer screening tools is a worthwhile goal. When MCED’s lifesaving promise becomes reality, it will be time for scientists, clinicians, policymakers, and the public to work together to make multicancer early detection succeed in practice.