Rates of colon and rectal cancer (CRC) incidence and death have been decreasing for decades. However, there’s an important exception to this trend. CRC incidence and death has been increasing for people below 50, all the way down to 20 years old. Risk still increases with age, but the difference in risk between 50-year-olds and 30-year-olds is smaller than it used to be. Younger people aren’t regularly screened, and doctors are often skeptical they can have CRC, so they tend to be diagnosed at later stages of the disease, which makes it more deadly.
In a 2016 statement, the United States Preventive Services Task Force (USPSTF) affirmed its prior recommendation that CRC screening begin at age 50. Just this year, they issued draft recommendations lowering that age to 45. They were beaten to the punch by the American Cancer Society (ACS), which lowered its recommended age to begin CRC screening from 50 to 45 in 2018.
Given that CRC deaths are on the rise for people as young as 20, should screening begin even earlier than 45?
Colon cancer background
First, a little background on CRC. The disease progresses slowly and in stages. One study breaks people down into seven stages of CRC that I’ve found useful while researching this post:
- Healthy
- Small polyp
- Large polyp
- Localized cancer
- Regional cancer
- Distant cancer
- Dead
Moving through the stages from health to dead from CRC can take 10 to 20 years. Most people with localized or even regional CRC don’t show symptoms. If the disease is caught in its localized stage, five-year survival rate is 90%. But if it’s caught in its distant stage, five-year survival drops to 14%. The confluence of all these traits makes CRC a prime target of screening, or testing people without any symptoms of disease. There’s a long period during which the disease can be detected and effectively treated but is asymptomatic. Once symptoms have started, it’s much more difficult to treat.
Colon cancer testing background
Next, a little background on CRC testing. There are two basic kinds of CRC tests, stool and visualization (aka structural). Stool tests are easy, cheap, and free of complications but less accurate. Visualization tests that look inside the colon and rectum are much more accurate but unpleasant, pricier (though, as we’ll see, still a bargain), and have a non-zero risk of complications. There are several variations of each kind, but I’ll focus on each category’s best test, a fecal immunochemical test (FIT) and a colonoscopy.
A colonoscopy is extremely accurate at detecting cancer and polyps. Plus, a doctor conducting a colonoscopy can remove polyps during the procedure, stopping the disease in its tracks. Because the disease takes so long to progress, after a colonoscopy that doesn’t find cancer, you can confidently go without further testing for years (exactly how many to be debated below). As effective as colonoscopies are, people hate taking them and a miniscule 0.003% of people who take them die from the endoscope perforating their colon. Stool tests are decent at catching cancer, but basically useless for catching polyps. If you get a positive result from a stool test, the next step is to get a colonoscopy.
A FIT looks for blood in your stool. It’s a huge improvement over an older test, gFOBT, that used different markers of blood in your stool. gFOBT could give false-positives depending on what you had eaten recently. Cologuard, a newer test, combines FIT with a test for cancer DNA in your stool. That sounds cool, and it does catch more cases of CRC on a single test, but it’s about 30 times as expensive and has more false positives. You’re better off taking a FIT once a year than one Cologuard every few years. See the table below, with descriptions of these tests from this paper.
Two quick definitions. “Specificity” is the rate at which people who don’t have CRC will test negative, a.k.a. true negatives. “Sensitivity” is the rate at which people who have CRC will test positive.
Recommendations from public health experts
The ACS and USPSTF both base their screening recommendations on what they call microsimulation models. These models take basic inputs like the information from the table above along with incidence rates and survival rates for CRC and then simulate what will happen to millions of people who progress through life under different testing regimes. The benefits they tally up include both years of life and quality of life. The costs include both financial costs and the unpleasantness and rare complications of colonoscopies. Each organization asked multiple academic groups to develop microsimulation models as a way of making sure that the conclusions from these models were robust.
In addition to the factors explicitly considered by the ACS and USPSTF, I believe that they both have an unstated but strong preference not to change their prior recommendations, which biases them towards expert opinions that were formed before we had good evidence. I will argue that without this preference, they would both recommend beginning CRC screening at younger ages than they do now.
For its 2016 recommendations, the USPSTF commissioned three microsimulation studies of CRC screening. They asked each of the groups to base their simulations on CRC incidence data from the late 1970s, before CRC screening at 50 became common. The goal was to avoid overreacting to an uptick in diagnoses of cancers that never would have been diagnosed otherwise because they wouldn’t have progressed to causing symptoms before the patients died of something else.
All three microsimulation studies found that beginning screening at 45 years old would increase life expectancy relative to beginning at 50, at the expense of additional colonoscopies over the average person’s lifetime. Two of the studies considered the possibility of lowering the first-screening age to 45 years but stretching the interval between colonoscopies from 10 years to 15 years. They both found that doing so increased life expectancy without increasing the number of lifetime colonoscopies. Depending on the difficulty and expense of colonoscopies, those results must mean that starting at 45 is better than starting at 50; the only remaining question should be how long to wait between colonoscopies. But, in 2016, the USPSTF chose to keep its prior recommendation to begin CRC screening at 50 years old with 10 year intervals.
Two years later, the ACS commissioned updated studies from two of the three microsimulation groups involved in the USPSTF recommendations. In particular, they asked the groups to reevaluate the benefit of CRC screening in people as young as 40 using more recent data on CRC incidence. Neither the USPSTF or ACS had ever recommended that people below 50 get screened, so the dramatic increase in the rates of younger people getting CRC most likely reflects a true increase and not just an increase in testing.
I dove deep into the weeds of the supporting information (SI) for the ACS microsimulation study. They considered 145 different testing strategies based on taking different kinds of tests starting and ending at different ages with different intervals between tests. In the SI, they showed that the strategy that maximizes life expectancy is nearly the most aggressive one they considered: colonoscopy every 5 years beginning at 40 and ending at 80. The only more aggressive strategy bumps up the cutoff age to 85 and ties on life expectancy, which adds 0.00 years of life.
Beginning screening at 40 does increase lifetime colonoscopies relative to starting at 45. The ratio of life gained to incremental colonoscopies comes out to 5.5 days. The ACS decided that’s not enough, and recommended starting CRC screening at age 45 but not before. Strangely, they recommend that “clinicians individualize screening decisions” based on “patient preference” for patients up to 85 years old. Their own study showed that extending screening on the low end from age 45 to age 40 is more beneficial than extending it on the high end from 80 to 85, but they recommend against the former while staying officially neutral on the latter.
In short, in 2016 the USPSTF found that moving the age limit down from 50 to 45 would help, but decided to stay at 50. In 2018, the ACS found that moving the age limit down from 45 to 40 would help, but decided to stay at 45. Last month, the USPSTF took a step towards matching the ACS recommendation.
Finding the minimum age at which CRC screening is effective should require, by definition, finding an even lower age at which CRC screening is not effective. Neither group has found that lower limit.
The benefit of extending colonoscopies from 45 to 40 is well within the FDA’s threshold for cost-effectiveness of hundreds of thousands of dollars per life year. I believe that these organizations are being slow to adapt to what their own evidence shows them out of a reluctance to contradict prior experts (who may be prior versions of themselves).
Hybrid FIT and colonoscopy strategy
The downside to additional testing is unnecessary colonoscopies. FIT screens are so cheap and easy the main cost to running them is that a false positive, which happens 3.4% of the time, leads to an extra colonoscopy. I expect you could maximize life gained with fewer colonoscopies by mixing FIT and colonoscopies. For example, the ACS microsimulations show that a colonoscopy interval of five years is superior to 10 years, to the tune of about six days of life gained per incremental colonoscopy. I would like to see a simulation of a strategy of giving a colonoscopy, waiting five years after a negative result, then giving FIT screens for four years until another colonoscopy at the 10th year from the first. I’d also like to see a simulation of giving FIT screens down to age 30 or even 25, followed by colonoscopies starting at some age.
FITs for millennials
I can’t reproduce these sophisticated microsimulation studies for younger patient cohorts. What I can do is use some simple Bayesian statistics as a heuristic. Both the ACS and the USPSTF agree that a 50-year-old, without regard to race, sex, or other risk factors, is at high enough risk to warrant a colonoscopy. I propose the following strategy as a supplement to the current recommendations: begin FIT screens annually at age 30. If the patient tests positive, immediately follow up with a colonoscopy. If the colonoscopy shows nothing, or only polyps that are removed, skip 4 years of testing, then resume FIT screens if below 45 or colonoscopies if 45 or older.
In the table below, you can see the average CRC risk by age for the average person, then for the same person after a positive FIT screen or a negative FIT screen. You can see that someone 25-29 with a positive FIT result is just as likely as an average person aged 50-54 without any test. If it makes sense for the average 50-year-old to get a colonoscopy, then it should make sense for an average 25-year-old with a positive FIT to get one, too.
I’ll acknowledge one factor I’m not considering: the psychological harm of a false-positive FIT. The risk of CRC is so low that even a 50-year-old with a positive FIT has a 98.8% of being CRC-free. And a 25-year-old with the same result is 99.94% likely not to have CRC. But the cost of a FIT is insanely low. You can buy one on Amazon for $25, and it’s even eligible for Prime. A two-pack costs $30, so you can save further by making your stool test into a romantic date with your significant other. The one I got had an expiration date 13 months away, so you might be able to use both yourself 12 months apart. The test takes less than 15 minutes start-to-finish in your own bathroom. Eliminating the 0.003% chance that a 25-year-old has undiagnosed CRC is a tiny benefit, but it’s cost-effective when weighed against the even more trivial cost of the test.
How to 115 