[groceryheist]

Sensationalist title to a good news story: there was no intent-to-treat effect, but a low conversion rate. The ETT is a pretty impressive 50% mortality reduction. Also study is limited by a short 10 year follow-up period but there will be a 15 year follow up.

So I'll still get my colonoscopy, thanks

[JamesBarney]

When everyone invited for a colonoscopy is compared to the control group 42% showed up and got a colonoscopy 18% fewer people got colon cancer same number of people died of colon cancer

When everyone who got a colonoscopy is compared to the control group 30% fewer got colon cancer 50% fewer people died of colon cancer

This data makes me think the mortality reduction benefit is bogus, but the cancer prevention benefit is real, and probably greater than 18%, maybe closer to 36-40%. If the colon cancer mortality benefit was real you'd see some reduction in the intention to treat group, and it'd be smaller than the cancer prevention effect. (The most aggressive cancers tend to be harder cancers to catch in time because they most so quickly, so most cancer screening tests will prevent more cancers than deaths)

[rippercushions]

Where are you getting "50% mortality reduction" from? The article says the reduction was zero:

> After 10 years, the researchers found that the participants who were invited to colonoscopy had an 18% reduction in colon cancer risk but were no less likely to die from colon cancer than those who were never invited to screening.

[tylermenezes]

> When the investigators compared just the 42% of participants in the invited group who actually showed up for a colonoscopy to the control group, they saw about a 30% reduction in colon cancer risk and a 50% reduction in colon cancer death

[Someone]

I don’t see how that says much about the usefulness of colonoscopy.

The people in the treatment group who didn’t show up must have had a 36% increase in colon cancer death (not explicitly stated in the article, but can be derived from the numbers. You need 42% × 50% + 58% × ? = 1), and (solving 42% × 70% + 58% × ? = 1) a 21% increase in colon cancer risk.

Something must have made them different from the control group. Maybe, they didn’t show up because they already were being treated for colon cancer?

[gwd]

Right, so to properly isolate the effect of the colonoscopy on mortality, we need to do a randomized trial with just people like you -- people who would definitely get a colonoscopy if invited -- to see what the effect is. Because you're also more likely to act on early warning signs and get a cancer diagnosis in time to treat the cancer rather than die.

[robertlagrant]

> Because you're also more likely to act on early warning signs and get a cancer diagnosis in time to treat the cancer rather than die.

This seems to imply we should discount all treatments because people who choose to get treatment are more likely to get better, coincidentally by the same amount as the treatment's efficacy.

[gwd]

That's not what it means at all.

The question here is a question that the medical establishment is trying to answer: namely, "What can we, as the medical establishment do, to reduce deaths from colon cancer?" For a long time, the answer has been, "Invite people to take colonscopies". What the data here appears to show is that the action, "Invite people to take colonoscopies" doesn't actually reduce deaths from colon cancer. If the medical establishment wants to actually reduce deaths from colon cancer, they'll need to figure out something else.

I guess I do agree that the headline is likely to be counterproductive. What the data might show is that the most effective thing you as an individual can do is to be the kind of person who takes colonoscopies when invited. The unfortunate effect it might have is to make more people into the kind of people who don't take colonoscopies when invited.

[robertlagrant]

> I guess I do agree that the headline is likely to be counterproductive. What the data might show is that the most effective thing you as an individual can do is to be the kind of person who takes colonoscopies when invited. The unfortunate effect it might have is to make more people into the kind of people who don't take colonoscopies when invited.

Agreed. Except the most effective thing you can do as an individual is have a colonoscopy, not be the sort of person who would have one :-)

[gwd]

> Except the most effective thing you can do as an individual is have a colonoscopy, not be the sort of person who would have one :-)

So we have two hypotheses here:

1. Having the colonoscopy is the thing that reduces deaths from colon cancer

2. Having the colonoscopy correlates to some other factor, X; and it's actually X which reduces the deaths from colon cancer.

X, for instance, could be a high willingness / ability to see the doctor when you experience early symptoms of colon cancer. That is, the more willing you are to go to the doctor when you start to have early symptoms of colon cancer, the more likely you are to survive it; and the more willing/able you are to go to the doctor when you have early symptoms of colon cancer, the more willing/able you are to have a colonoscopy.

What evidence do you have to believe that #1 is true, rather than #2?

Because if #1 is the case, the medical system should push hard on colonoscopies. But if #2 is the case, pushing colonoscopies might be a red herring. In fact, it might be counterproductive -- I've heard that colonoscopies are unpleasant; if you pressure people who don't like doctors into having a colonoscopy, and they have a terrible experience, then when they experience early symptoms of colon cancer, they may be more likely to procrastinate to avoid having another one. Rather, if #2 is the case, the medical system should try find out what can be done to make people more willing / able to get early medical care.

[vidarh]

No, but you need to account for it when considering efficacy, because people who choose to follow through on treatments may also be taking other steps that may also improve their outcomes, and it can be hard to tease out because the mere act of telling them they should have a screening might change other behaviours even without being prompted by doctors in a way that's recorded.

E.g. it's a reasonable hypothesis that patients who are more motivated to show up might also be more motivated to look up possible causes and what other steps they can take to improve their chances.

In other words, it's reasonable to expect people who comply to potentially get better at a higher rate than the efficacy of a single treatment, and teasing out how much of this effect is due to the intervention itself and how much is due to changed behaviour due to the referral is hard.