Study-of-the-Week: One Last Comment on the "Great Colonoscopy Debate"

A medical conservative moves a bit toward the center

My sincerest apologies for dragging “The Great Colonoscopy Debate” into a second week. Given that Drs. Mandrola and Prasad tried to gang up on me in a debate, and both of Sensible Medicine’s posts last week (Monday and Thursday) took one side of the debate, it seems necessary to present the other side. This article will be a bit long and goes a bit into the weeds. Bear with me, this is an important topic, there are excellent arguments on both sides, and we are Sensible Medicine, not afraid to be thoughtful and entertain opinions with which we might not agree.

I come to this debate as someone quite skeptical of screening. I am an avowed medical conservative who thinks that a doctor’s role is primarily to treat disease.[i] Anything that has the potential of turning a healthy person into a patient should be examined in a very bright light. I have always argued that the evidence for a screening test should include the following:

• A randomized controlled trial comparing people who are offered screening to those who are not.

• Demonstration of a benefit in overall mortality. It does not matter if prostate cancer screening reduces deaths from prostate cancer if those screened die at the same rate as those not screened – either because of their age or because of the off-target harms of screening.

• An intention to screen (ITS) analysis. I won’t go into great detail on this (Dr. Prasad wrote about screening trials in depth in chapter 4 of Ending Medical Reversal). Suffice it to say that ITS maintains the randomization that you worked so hard to achieve, most closely mimics the real world, and biases your results to not finding a difference between the two groups (the more conservative approach).

It is based on data from studies that fulfill these requirements, and I am circumspect when I counsel patients about breast cancer and prostate cancer screening.

Let’s get into last week’s NordICC study published in the NEJM and why this article demands a particularly close read and the ability to think outside the “screening trial appraisal” box a bit.

Looking at this study at an “Abstract” level, you walk away with the impression that colonoscopy-based colon cancer screening is ineffective. The study is an impressively large, population-based RCT conducted with an ITS analysis and reporting overall mortality. The death from any cause was essentially identical in the two groups during the 10-year follow-up period: 11.03% vs. 11.04% (0.99; 95% CI, 0.96 to 1.04). Even the colon cancer-related death rates were the same: 0.28% vs. 0.31% (0.90; 95% CI, 0.64 to 1.16).

If you wish, you can stop there. But if we’re a little more thoughtful about the results, I’ll argue that there is really good reason for a person to leave this article with robust evidence to support a decision to get a colonoscopy. Let’s dive in, 3 points, and then a second look at the data.

1.      Colonoscopy is different from other screening tests. Mammograms, PSA tests, CT scans for lung cancer work (if they do) by finding cancer early, when it is more treatable – and remember, we don’t screen to find cancer, we screen to save lives.

Colonoscopy is different. Colonoscopy actually prevents cancer. A precancerous polyp that is removed poses no threat of mutating/growing into a cancer. Preventing cancer is a big deal. I am not sure anyone would rather have cancer treatment than a colonoscopy with polypectomy.

2.      As has been discussed in numerous places, this was a study of a national screening program based on invitation for colonoscopy. The uptake was low. I am a primary care physician. I have recommended colon cancer screening (colonoscopy being the preferred method) for years. I get data on my own performance. Upwards of 80% of my patients who are indicated have had a colonoscopy. An uptake of 80% vs. the 42% observed in the study would have affected the results.

3.      I do not think the ITS analysis is necessarily the most important one in this study. The alternative to ITS is a “per protocol analysis”. In a per-protocol analysis, you compare people not by the group to which they were randomized (ITS) but by what treatment they had. The problem with per-protocol is that your screened group might be very different from those who go unscreened – the two groups are no longer random samples, and you have introduced confounding. Your screened group may be healthier, wealthier, and wiser (or anxious, masochistic, and idle in this case).

In this study, however, other factors that would have made you more or less likely to benefit from screening would have either made you ineligible for this study or would have been very weak risk factors. I think it is unlikely that confounding played much of a role. Also, the researchers did yeoman’s work controlling for confounding—see paragraph 3 of the “statistical analysis” section of the article. Saturday’s article on Sensible Medicine also looks at adjusting the per-protocol analysis using instrumental variables.

Taking these three points, let’s look at the results of the study again.

If we are looking at colon cancer prevention, colon cancer was reduced in the ITS analysis from 1.20% to 0.98% (0.82; 0.70 to 0.93). In the per-protocol analysis, these numbers go to 1.22% to 0.84%, (0.69; 95% CI, 0.55 to 0.83). Yes, these are a small absolute reduction,s but it is a 31% risk reduction of a colon cancer diagnosis. This is a big deal. 31% risk reduction is orders of magnitude larger than anything else we have (aspirin, fiber, turmeric…). And this was done with essentially no harm – colonoscopy was shown to be exceedingly safe. For those of you who like the “number needed to screen.” These numbers equate to a number needed to screen of 263 to prevent one colon cancer. For a patient, there is a 1/263 chance that their colonoscopy will prevent cancer. For a doctor, for every 263 patients you refer for a colonoscopy, you prevent one colon cancer.

To some people, these numbers will seem huge. “Get a colonoscopy when it is that unlikely to help me? Forget it.” For others, this will be a convincing argument to take the test. Sensible Medicine commenters love to take me to task (appropriately) when I add an anecdote, but I’ll take my chance here. In my years as an internist, I have watched 4 of my patients die of colon cancer. The only thing they had in common? They did not have colon cancer screening. Among my screened patients. no deaths.

Now, I still have lots of reservations. Although I am pretty sure that colonoscopy is the best colon cancer screening/prevention tool we have, it is not a great test. It requires an unpleasant prep, anesthesia (for most people), and a day off from work. The recommended age for initiating screening and the screening frequency are not based on robust data. The cost-benefit is something that truly needs to be debated.

This study reminds me that colon cancer screening is a decision, like all healthcare decisions, to be made by a patient in consultation with his or her physician. Some may choose to partake in colon cancer screening. Some may decline. Patients should not be shamed for their decision, in either direction. Doctors should not be judged (or reimbursed) based on how successful they are in referring patients for colon cancer screening.

---

[i] This is an unpopular stance in this country where we are raised with the idiom, “an ounce of prevention is worth a pound of cure.”