Does the NordICC Trial Dethrone Colonoscopy for Colon Cancer Screening?

A first-ever RCT on the benefits of colonoscopy arrives, and it underwhelms.

It’s not often in the world of cancer screening that a study comes out which merits a description like “bombshell.” This week, however, we got one: the NordICC randomized controlled trial on the benefits of colonoscopy for colorectal cancer screening. To the surprise of many (and the dismay of most gastroenterologists), no meaningful benefit was found. Does this mean we should stop recommending colonoscopies to screen people with average risk?

NordICC doesn’t fully answer that question, but I do think we need to mull over what we’re doing a little harder.

First, though, it’s important to think about why we screen for cancer in the first place: to catch it early enough to make a difference in treatment outcomes. A small, aggressive breast tumor caught early by a healthy woman’s mammogram which can be removed by a lumpectomy and treated with excellent results via a well-tolerated hormonal agent is the ideal screening “save.” In contrast, a “positive” mammogram might also involve finding an indolent breast tumor which might not have ever progressed but led to surgery and chemotherapy in a woman with advanced cardiac disease likely to die of a heart attack before her breast cancer would have impacted her health. Contrary to what the purveyors of “executive physicals” will tell you, it’s not always best to know every little collection of cells inside our bodies.

Evaluating cancer screening tools is a tricky business. Imagine looking only at the number of cancers detected with and without a screening program. If I order a PSA every year on my 85 year old male patients, I will find a lot of prostate cancer compared to if I follow published guidelines and don’t go looking in the first place. However, the cancers I find are unlikely to help my patients live better or longer. My little screening program might seem like a success because I am “finding” a lot of cancers, but really I might only be increasing cancer diagnoses without improving outcomes like deaths from prostate cancer or all-cause mortality — the problem of “overtreatment.” I think of an elderly patient of mine with substantial heart disease whose urologist kept checking PSAs annually; when they rose, he biopsied him; and the biopsies came back positive, leading to a definitive diagnosis of prostate cancer and a good deal of angst and discussion. However, when his heart disease risk and age were taken into account, treating the prostate cancer would not have improved his ten year survival, so the urologist went back to watching PSAs. A great win for cancer screening!

However, if the actual number of deaths attributed to a given cancer is lower in the group being screened than the unscreened cohort, that counts as a meaningful improvement. Best yet is to find a reduction in all-cause deaths; this way, we don’t need to worry that things like all the extra drives to the hospital to follow-up positive mammograms, the extra infections from chemotherapy, and additional drug-drug interactions from treatment regiments, don’t end up leading to just as many deaths as if the cancer had not been detected early.

Few are the studies of cancer screening tests that have found all-cause mortality benefit. (For this reason, I found the shock that many preventable disease deaths actually decreased during the 2020 lockdowns rather amusing; I know all y’all think we doctors are life-saving Heroes, but unless you are bleeding out or having a heart attack, we often do more harm than good). Reducing deaths from all causes might be too high a goal; after all, most people of screening age don’t die in a 10 year period, and when they do, it is usually not of cancer, not to mention only one type of cancer. From Our World in Data, the many causes of death in 50-69 year olds and types of cancers leading to death:

It’s hard to move the all-cause mortality needle to a statistically significant degree when focused on one cancer type, barring remarkable results.

Timing is another sticky issue in cancer screening; namely, lethal cancers tend to progress more rapidly and are harder to catch with periodic screenings, while slower-growing cancers are easier to catch. One solution is to simply increase the frequency of screenings, but of course that carries with it two new problems: increased cost, since we don’t have an endless supply of health care spending and need to allocate resources to the areas which most improve quality of life; and adverse events, since nothing in life comes without risk. In the case of cancer screening, most of the risk lies in potentially unnecessary, unpleasant and possibly dangerous medical procedures when the screening test is positive. Some 90-95% of mammogram “positives” are false positives, meaning the additional images, needle biopsy or surgical excision that followed represent a highly stressful wild goose chase.

In the case of colonoscopy, there are two unique factors to consider. The procedure itself carries with it some real, albeit small, risk; several studies have found that 1-2% of people undergoing colonoscopy end up in the ER or hospital for complications soon after the procedure. (I’ll note that this in no way passes the eye test for me, having had probably thousands of patients undergo a colonoscopy and certainly not recalling more than a couple who ran into trouble with their electrolytes or severe abdominal pain, and none experienced a bowel perforation, akin to the experience of the NordICC study participants.) However, there is also a remarkable benefit to colonoscopy, and the reason why I am often encouraged to recommend this “gold standard” procedure: unlike a PSA test or mammogram, it’s possible to snip pre-cancerous lesions out of the colon during endoscopy, thereby theoretically not just screening for colon cancers but actively preventing them. “Preventative screening” — Pap smears would be the other example of this — has a certain pro-active ring to it.

That’s an ideal screening tool: one that can alter the disease course for the better and prevent the need for treatment at all, as well as pick up cancerous lesions early enough to respond to less invasive treatments, and when prognosis is still good.

If that were the case, we might expect that while people screened with stool tests alone, which look for evidence of microscopic blood escaping more advanced polyps or cancerous lesions (the heavily advertised Cologuard test also looks for DNA markers of colon cancers), would have their colon cancers detected, there would be more of them compared to a group getting regular colonoscopies, since they would not be getting their colons pruned of pre-cancerous lesions periodically. This should result in fewer actual cancers, better colon cancer survival rates, and, possibly, improved all-cause mortality. This would be the holy triad of colon cancer screening!

Therein lies the great interest in this study. While screening via stool testing has been shown to reduce the rate of death due to colon cancer, it’s not been shown to reduce all-cause mortality. Flexible sigmoidoscopy, in which only the first third of the colon is explored endoscopically after a milder colon prep than full colonoscopy, when offered every 5 years has been shown to provide a modest overall mortality benefit. Surely, then, exploring the entire colon must at least match this benefit?

It didn’t, though, in NordICC, and that’s what has set the gastroenterology world on fire.

For the study, about 28,000 people between 55 and 64 in Poland, Norway and Sweden were randomized to be invited to undergo colonoscopy and followed on average for 10 years, while about 56,000 served as the “usual care” controls. (“Usual care” in those countries implies zero colon cancer testing; apparently these frontier nations do not love their citizens enough to even try to prevent them from dying en masse from colon cancer.) Only 42% of invitees actually got a colonoscopy. Of that whole group, there was no overall mortality benefit, and only a very modest reduction in colorectal cancers detected — roughly 1% vs 1.2% — and even more modest, and not statistically significant, reduction in colorectal cancer deaths, 0.28% vs 0.31%.

These were fightin’ figures. Responses were prompt, and sometimes professional:

Sometimes indignant:

And sometimes funny, in a pointed sort of way:

All of these are fair points. In a pharmaceutical randomized controlled trial, we expect only a tiny proportion of participants randomized to the test group to opt out of taking their pills or getting their injections. After all, they already signed up for it! A trial involving one-time invitations for an unpleasant prep and an invasive procedure makes for low participation numbers, which deeply affects how well the data can transfer to the general population. I’m pretty sure I get higher than a 42% rate of acceptance for colonoscopy when I recommend it to a patient in person.

I take the other concerns seriously, too. Maybe less-practiced Scandinavian and Polish gastroenterologists aren’t as skilled at endoscopy as their American counterparts, although the endpoints, like the rate at which they detected adenomas and reached the far end of the colon, did not seem poor enough to me to skew the results greatly.

Most importantly, when studying only those who actually bothered to get the screening colonoscopy, the authors found approximately 50% reductions in colon cancer cases and deaths — although the drop in absolute terms was small, from about 1.2% to 0.8% for cases, 0.3% to 0.15% for deaths. And, as Hematologist/Oncologist gadfly, Dr. Vinay Prasad, points out, shifting the outcomes from the randomized intention-to-screen group to only those who chose to undergo colonoscopy blows the benefits of randomization, which really is why the trial was so important in the first place. After all, we have earlier data, like this study from Kaiser, showing a substantial (67%) reduction in colorectal cancer death among those who received colonoscopies, but it’s retrospective observational data, prone to confounding biases like, “Maybe the people who don’t mind a night of diarrhea and having a camera passed through their anus are different than the rest of us when it comes to health care?”

It’s possible to argue both sides of this study, and it’s not “bothsidesing;” this is an important study despite its flaws. Those really interested in picking NordICC apart can devote an hour to this discussion; internist Dr. Adam Cifu calmly and patiently resists the temptation to verbally strangle his old friend, Dr. Prasad, while making the point that colonoscopy remains a valuable preventative screening tool for colorectal cancer despite this one study.

My take-home from spending quite a few hours following the debates over this study is that my pitch for colonoscopy should be more nuanced. I used to tell patients, based on prior observational data, that if they are 50 (now 45, per the new recommendations!) and in average health, they might have a 4% risk of developing colon cancer over their lifetime, and about half that risk of dying of colon cancer. Screening with colonoscopy, I reckoned, should cut those numbers roughly in half, while annual Fecal Immunochemical Tests (“FIT”), which have replaced stool guiaic testing in our state, might be three quarters as effective as colonoscopy.

Now? I’d temper those figures a bit. Given the sub-analysis of NordICC patients receiving colonoscopy vs the entire randomized cohort, I might frame those reductions in the “20-50% ballpark” or “maybe decreased by about a third.” If the 10-year benefit from a colonoscopy now drops the risk of colorectal cancer from 1.2% to 0.8-1%, a 55 year old average risk patient is looking at 250-500 lifetimes for the one decade of life which benefits via a prevented colon cancer. All those other hundreds of lifetimes have a guaranteed nasty prep, some missed time from work or family, and maybe a 1% risk of getting sick from the procedure. It’s not a no-brainer, on the individual level.

On the societal level, there are a lot more saves, but a lot more expense. If colonoscopists charge Medicare $1000, conservatively, that’s a lot of health care expense for the 80 million Americans between 45 and 75. For the fiscally responsible (and, perhaps, unempathic), if colon cancer costs in the $60,000 ballpark to treat, a Number Needed to Screen in the several hundreds does not amount to a financial winner.

This is especially true since we have an inexpensive option, the FIT stool test, which is currently being tested against colonoscopy in a Swedish trial. At $20/test done annually, it also stacks up nicely against its $600 Cologuard cousin done every 3 years; especially since we have none of this sort of randomized controlled trial evidence for Cologuard, just a hope that its modestly higher sensitivity (which comes with a higher false positive rate, alas) would amount to it being a more effective screening tool. I have my doubts.

We don’t have — yet — randomized controlled trial data for the FIT test’s effect on colon cancer mortality. I can say that what I like about annual stool testing, aside from the low cost, is the frequency. Yes, it will miss some cancers with its lower sensitivity, reportedly in the 70-80% range per test vs colonoscopy’s 95+%, and miss most pre-cancerous lesions, but perhaps this is more than made up for by running the test every year rather than waiting 10 years.

We do have RCT data for flexible sigmoidoscopy, and it’s good. Again, the point of view that we are missing out on screening the 2/3rds of the colon where some 30% of colon cancers develop and therefore it simply must be inferior to full colonoscopy is based on common sense, but sometimes common sense fails in medicine. Flexible sigmoidoscopy — which everyone in the medical universe refers to as a “flex sig” — is often performed by general practitioners and is a fraction the cost of colonoscopy, not winning it much love from the gastroenterology community but making it popular among single payer health care systems abroad as well as patients who would rather have a milder prep and less invasive experience twice a decade rather than a full prep and deeper anesthesia once. As to why there might be better data for survival benefit for a flex sig over colonoscopy, perhaps it really all comes down to higher participation rates in the flex sig studies (generally over 60% vs 42% with NordICC). But maybe, just maybe, the real key in colon cancer screening is frequency rather than thoroughness. Perhaps the shortcomings of a sigmoidoscopy might be more than made up for by screening twice as often.

The thought process here is that different colon polyps and cancers grow differently. It’s hard to study this accurately in humans, for obvious reasons; best of luck enrolling a trial of Christian Scientists who refuse treatment of their colon cancers but are game for monthly colonoscopy to follow the progress of their tumors. A small study based on CT imaging of cancerous colorectal lesions that were not being surgically treated (granted, hardly representative of the average person showing up for a screening colonoscopy) showed a huge range in the time tumors took to double in size, from under 4 months to well over a year. We have theories, but don’t even really understand what causes some colon polyps to form into invasive cancers, while the majority regress or don’t continue to grow. All of this makes it difficult to fathom which screening tool at what frequency will minimize false alarms while still finding potentially lethal cancers.

I am not ready to declare that, despite the obvious appeal of colonoscopy to visualize the entire colon and prevent polyps from developing, the NordICC study convincingly shows it is not the test of choice. That 42% participation rate was just too low to take these unimpressive risk reduction numbers at face value. However, the modest benefit even when constrained to those receiving colonoscopy raises some questions about the utility of colonoscopy for individuals, and its societal cost-efficacy. With the absolute gain so small against no screening (“usual care”) whatsoever, it presumably would have shrunk dangerously close to zero against an alternative screening program like FIT testing.

I’d like to say, “let’s wait for the Swedish data to decide,” but it’s a 15-year trial started in 2014, and while benefitting from a head-to-head comparison of colonoscopy vs FIT vs usual care, there will still be doubts about the colonoscopy arm if participation is low. That, unfortunately, seems likely, given that once again it was based on a mailed invitation with anticipation of low compliance rates:

On balance, what I take from the NordICC study is that we should ask more questions about the value of colonoscopy. Perhaps it is best at finding the “wrong” cancers and pre-cancerous polyps — slow growers that would have been picked up with less sensitive screening methods — and misses the fast growing cancers that kill us because checking once a decade just isn’t enough. Maybe it’s great at detecting lethal colorectal cancers, and was sand-bagged by a study with unfortunately low participation rates. I don’t know, but can say that my needle has moved a little away from colonoscopy while awaiting more data.

So, in an imperfect world, what is a family doctor to recommend? Lacking evidence that colonoscopy exceeds the benefits of annual stool testing, the FIT test will continue to be my recommendation for normal risk patients, although I would happily offer flexible sigmoidoscopy if it were available here, for average risk patients for whom “half the hassle, twice as often” has an appealing ring. Those with strong family history of colon cancer, inflammatory bowel disease, or the fairly rare genetic diseases predisposing towards colon cancer, have much higher absolute risk of developing colon cancer and continue to very likely benefit from regular colonoscopies. For the rest, though, I favor the approach of many organizations, and entire nations like Canada, which is to eschew colonoscopy in favor of non-invasive screening.

Now, while I believe most of the outcry against the NordICC trial to be honest and well-intentioned, my Inner Cynic observes that there might be a trillion reasons for such strenuous objections from the gastroenterology community. Per a 2020 Medscape article:

…

OK, I went and said it, and now my hopes of getting a Christmas card from our local gastroenterologist are probably shot to pieces. I’m sorry, Ron. I’ll still need you for all the positive FIT tests.