The Coming Revolution in Cancer Screening

The next decade should bring a revolution in early detection of cancer. Today, only four cancers have screenings recommended for the general population: breast, colon, cervix, and lung cancer in smokers. Those cancers only account for 29% of all cases of cancer and 24% of all cancer deaths. Due to low compliance with recommended screenings and some cases of cancer missed by screening, only 15% of all cancers are caught by currently recommended screenings.

Given such low rates of early-stage cancer detection, there’s a large opportunity for improved screening. Enter liquid biopsy blood tests. Tumors often shed DNA into the bloodstream, known as circulating tumor DNA (ctDNA). Blood tests that detect ctDNA can detect many different cancers at an early stage, before they would appear on imaging tests. 

Galleri

The most well developed of these tests is Galleri, which can detect over 50 cancers. The test is already sold as a lab-developed test. GRAIL, the company that makes Galleri, would like everyone to take the test once a year from ages 50 to 79, in addition to existing recommended screenings. Even if the test is able to detect many cancers, there are two concerns with recommending that so many people take it. First, false positives: people may get positive results for cancer they don’t have. Second, overdiagnosis: people may find out they have cancer that otherwise would never have caused a problem and pursue expensive treatments with harmful side effects. 

The first major study of Galleri, the Circulating Cell-free Genome Atlas (CCGA) demonstrated the test was able to detect 44% of stage I-III cancers. Although missing over half of cancers sounds bad, that would more than triple the fraction of cancers detected by screening. If everyone 50 to 79 years old got the test annually, 49% of all cancers would be detected by screening, up from 15%. And the flip-side of missing so many cases of cancer is giving very few false positives: only 0.7% of people without cancer incorrectly tested positive. Combined, that means that for every person who is correctly diagnosed with cancer, 1.8 people would get a false positive. In comparison, with existing screenings, for every person correctly diagnosed with cancer, 43 are given false positives. Added to existing screenings, that’s 14 false positives per correct diagnosis, still far fewer than today’s 43.

To look into the problem of overdiagnosis, GRAIL published up to three years of follow-up for 2,000 patients from the CCGA trial who were diagnosed with cancer. All of them found out they had cancer through existing methods, meaning current screenings or going to the doctor’s office with symptoms. Cancers the new test detected were much deadlier than cancers it missed, which suggests it may not add much overdiagnosis.

Cost effectiveness

How much should we be willing to pay for the new test? GRAIL’s own analysis – surely biased, but worth considering – finds that if we’re willing to pay $50,000 for an additional year of healthy life, we should pay $904 for the test. GRAIL’s estimate is built on simulations. Using data from the CCGA for how many cancer cases the test detects at what stage, and historical data for survival rates versus stage at diagnosis, the company estimates that taking the test annually adds just under 0.2 years of healthy life. They weigh that against the costs of additional testing after receiving a positive Galleri test, as well as the cost of treating any confirmed cancers. The assumed willingness-to-pay of $50,000 per year is conservative: federal agencies typically assume a value closer to ten times as much based on how much extra pay workers demand to accept dangerous jobs. 

The best way to measure the life extension benefit of the test is with a randomized controlled trial (RCT). So the UK government is starting a gigantic RCT with 140,000 people, with plans to add one million people if early results are positive. Patients will have annual tests for three years. Half will get their test results immediately. Samples from the other half of patients will have their samples stored until the end of the trial. The experiment should tell us what fraction of late-stage cancer diagnoses are shifted to early-stage and how much overdiagnosis results. First results are expected in 2023. If those look good, the trial will add a million patients over the next two years. Measuring the impact on cancer mortality will take longer – there’s a time delay between diagnosis and death – so final results may not come until 2030 or later. In the meantime, every two or three years we should get meaningful updates. 

To test or not

As a 32-year-old, I have plenty of time to wait on the RCT results. If I were in the recommended age range of 50 to 79 years old, I think I’d ask to get the test if I had the $950 to spare (not medical advice!). The company’s own analysis of cost effectiveness is compelling and the fact that the UK NHS is already gearing up to roll the test out to so many people is a major vote of confidence. 

Whatever we do today, in a decade there should be about twice as many people walking around with an early-stage cancer diagnosis. The difference will be much larger for cancers that currently have no screening and are difficult to detect. In particular, pancreatic cancer, which killed my grandfather, is the third leading cause of cancer death, just a hair behind colon cancer, but is rarely detected in the early stages. Hopefully, once many more people receive early-stage pancreatic cancer diagnoses it will become easier to run trials of new treatments and fatality rates will improve at an accelerating pace. 

Leave a Reply