One Physician’s Thoughts On That Colonoscopy Study, And The Debate That Ensued

Many people took umbrage with the study I posted this week that found, essentially, screening colonoscopies don’t save lives. They HAVE to, right? Early detection, right? That’s not what science is finding. Here’s an east coast physician addressing the study and the debate that ensued.

Great Colonoscopy Debate – A Listener Responds, Sensible Medicine, 13 October 2022

After watching the great debate last night, I had 5 thoughts.

(1) The point Dr. Mandrola made about the actual numbers of the per-protocol analysis got obscured. I don’t think people who hadn’t read the article realized that the risk of colon cancer-related mortality was 0.30% in the usual-care group and 0.15% in the intervention group. All-cause mortality wasn’t even reported for the per-protocol analysis.

Even those who embrace per protocol analysis are making a losing argument. We’re talking about an absolute risk reduction for cancer-related mortality of 0.15% even in the best-case scenario (per-protocol analysis). I’ve shown it here visually against the backdrop mortality.

The 2 tall bars on the left show the 10 year risk of dying seen in the study. The third bar assumes 100% compliance with screening, and the short bars are rates of colon cancer death observed (4 and 5), or rates of death assuming 100% compliance (6 and 7). Simply put: colon cancer death is dwarfed by dying for any reason, and the benefits are tiny.

(2) Where are people getting the number 50% reduction from for the per-protocol analysis with regard to cancer-related mortality? This is being parroted by all the national GI organizations. It’s from this 0.30% and 0.15%. Dr. Mandrola was also trying to make this point, but it got slightly obscured in the discussion. Relative risk obscures the very low absolute benefit.

Here’s the kicker: All the hot-take commentators, of which there are many out there, don’t realize: they are parroting the NEJM editorial which quoted the 50% reduction in cancer-related mortality in the per-protocol analysis.

I am very certain the editorialists wrote it like this in the editorial rather than provide the actual raw numbers published in the paper because they knew this could become a defensive talking point for people who wouldn’t bother to read or understand the whole paper.

Yes, I am that cynical about academic medicine.

And yes, I am fairly certain this framing was intentional.

(3) The description of the trial in the Methods was that they expected a 50% participation rate in the intervention arm. 42% is lower than this, but it’s not dramatically off from what they had expected. This has not been mentioned.

(4) If the issue of colonoscopy saving lives is settled already, why was this trial even done? We are not doing trials to determine if the Earth is round because that is considered settled knowledge. We do trials to help with clinical issues which are not settled. If the trial had happened to show a positive finding, the same people getting worked up about methodological issues now would be championing the study — I am very sure of this. The truth is there are 3 more ongoing trials precisely because this is not settled science. We have no idea which colorectal cancer screening is best in 2022.

(5) I have never observed so many trainees (GI fellows and IM residents who want to do GI) get so worked up over a scientific issue on Twitter. I am sure that >80% of them don’t even understand the nuances of the discussion. Some are doing it just to demonstrate they are on the “right team” and advocating for the right cause (let alone the obvious COI that their salaries are highly dependent upon this approach to medicine).

Their ability to analyze unemotionally is entirely gone. They know the answer they want to arrive at and see no nuance whatsoever.

Good points.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s