[aydyn]

> explained that if breast cancer caused 3% of all female deaths and screenings reduced these deaths by 35%, that’s a good result on its own. But screenings may change mortality overall by only about 1%

This statement makes it clear what's going on. In ML terms: class imbalance. 99.9% of people won't ever get colon cancer, and therefore won't ever benefit from a colonoscopy. It won't make any statistical difference in overall population survival. But for the 0.1%? It will save their lives.

[vc8f6vVV]

> 99.9% of people won't ever get colon cancer, and therefore won't ever benefit from a colonoscopy.

But they may have complications from a colonoscopy, that's the idea. No test is completely harmless, even a blood work. You save some lives but may loose others, that's the point of the paper. And of course you waste resources that can be used to find a cure.

[ilya_m]

I understand the comment's thesis, but I'd like to see more accurate numbers used where real health outcomes are concerned.

The lifetime risk of being diagnosed with colorectal cancer is ~4%. (With the odds trending higher for the younger generation.) The risk of _death_ from this cancer is ~1%.

[hedora]

Yes, and if you perform dozens of tests per year on the general population, and take action based on the results, then the probabilities add up pretty rapidly.