By Dr. Ronald Adler, Associate Professor, University of Massachusetts Medical School
In 2016, the DNA-sequencing company Illumina made headlines for launching GRAIL, a health care start-up with the goal of creating a blood test that can detect all types of cancer before symptoms arise. In the following years, GRAIL and many other companies have presented initial results of tests that detect the DNA of cancer tumors in the blood, growing the anticipation for usable tests.
Last week, GRAIL’s chief medical officer Dr. Josh Ofman presented an update on their work for the American Society of Clinical Oncology (ASCO) Annual Meeting, ushering in what he calls a “New Era of Early Cancer Detection.” Many aspects of this project seem impressive: Their blood test can reportedly detect more than 50 cancers, and identify their location in the body, with a false positive rate of less than 1%.
Yet there is still reason to be wary of the promises of GRAIL and other “liquid biospy” companies. GRAIL’s mission is based on the belief that detecting cancer early leads to cures. However, while this is true for some individuals and despite the popularity of the phrase “early detection saves lives,” this has not been clearly shown for most cancer screening programs (with the notable exception of cervical cancer screening). At best, some of our cancer screening programs have demonstrated cancer-specific mortality benefits, but not all-cause mortality benefits. The reasons for this are numerous, but the key contributors are improved cancer treatments, which render early detection less impactful, and the fact that harms associated with screening (testing complications, false positives, overdiagnosis, and overtreatment) typically offset the benefits in a population.
While GRAIL’s low false positive rate is encouraging, we should also be concerned about the downstream effects of the true positives. Not all cancers are equally dangerous. Some types of cancer will never grow or cause harm (turtles), some are so aggressive that they are fatal no matter how early they are caught (grenades), and some cancers could be lethal but also treated successfully if found early (bears).
The ideal cancer screening program would help us find only the bears, because the point of cancer screening is to save lives through early detection, not merely to find more cancers. But so far, our cancer screening tools have been inadequate at distinguishing between these types of cancer. Research from the past several decades has shown that for many types of cancer, screening has increased the incidence (how much cancer we find) but not changed mortality (how many people die of cancer). To the extent we have seen reduced cancer mortality rates, most experts agree that these are mainly due to lifestyle changes and improved cancer treatments, not screening.
Dr. Ofman also falls into the trap of lead-time bias by citing five-year survival rates as evidence of screening benefit. He writes,”a person whose cancer is found before it has spread to other parts of the body is more than four times as likely to survive five years than a person diagnosed with advanced disease.” Some cancers, the indolent “turtles, would never cause harm and are destined to never spread, so five-year survival rate is not a useful measure of the value or utility of a screening program. A patient could be diagnosed with cancer earlier in the course of their disease, live for five more years, but still die of the cancer at the same age as someone diagnosed later.
To properly judge the value of a screening program, we need to look at mortality rates, and careful consideration needs to be given to the heterogeneity of cancer. Finding cancers for which no treatment is needed (turtles) or cancers for which we have no effective treatment (grenades) does more harm than good. A true advancement would be a test that helps us distinguish for example, an indolent prostate cancer from an aggressive one. Can their test do that?
Finally, we need to honor the voices of our patients. Though this is “only a blood test,” it represents a high-risk procedure, because the stakes are very high. Diagnosing patients with grenade- or turtle-type cancers transforms them from people to patients with no mortality benefit and a great deal of associated physical, psychological, and financial harms.
There has been a growing movement towards prioritizing shared decision-making (SDM) before proceeding to elective procedures such as cancer screening. This is now considered standard of care for prostate cancer screening, and Medicare requires it for lung cancer screening. A Cochrane review has demonstrated the utility of decision aids for facilitating effective SDM. (For a primer on SDM, a review of cancer screening, and a variety of decision aids, consider Cancer Screening Decisions: A Patient-Centered Approach.)
Patients should be fully informed of the balance of risks and benefits before agreeing to such testing; if properly informed, it is likely that many patients would decline such a test. Perhaps more importantly, such testing should not be approved until mortality benefits are demonstrated.