Hacker News Reader

Comment by nluken | Hacker News Reader

I get what you're saying: it's a bit of an apples-to-oranges comparison since better screening means we're comparing cancers of different severities. I don't necessarily see that as a problem, though, since a lot of these earlier screened cancers would have become more advanced cancers down the line. If our testing were perfect such that we could detect all cancers early enough to stop, that would functionally be a cure, no?

View on HN

The "US Cancer Death Rate" talked about in the article is measured in deaths per 100,000 population, not deaths per detected cancer. So we are comparing apples to apples, but finding and treating cancers earlier could be helping to result in fewer deaths.

Or it could simply mean that another cause of death has increased. Like heart disease/diabetes from ever increasing obesity rates.

A person can't die twice, so if heart disease gets them first, then they won't die of cancer.

> Or it could simply mean that another cause of death has increased. Like heart disease/diabetes from ever increasing obesity rates.

You could hardly have picked a worse counterexample. Deaths from heart disease have fallen dramatically over that time period, even faster than cancer deaths have dropped.

Deaths from strokes - which at one point equalled deaths from all cancers combined - have fallen as well, although much of that drop occurred before the time period in question.

Deaths from diabetes are actually quite few in number compared to heart disease and cancer. They've sort of plateaued over this time interval, which isn't good news, but hardly the alternate explanation for the reduction in cancer deaths.

Now, there is one big cause of death which we haven't addressed: COVID-19, which causes about 3x the number of deaths as diabetes, and over half the number of deaths as all cancers combined. But that only began in 2020, and this data shows a decrease across the entire interval, so it's not relevant here.

Diabetes is almost never the direct cause of death, but it’s damn hard in the body. Hurts circulation, hurts your nerves, hurts your organs… and it can flare badly in reaction to some medications.

So, the cause of death won’t be “diabetes”, it’ll be heart attack, stroke, chronic infection, something like that.

Similar to AIDS in that it doesn’t kill you directly (not that we know that juvenile diabetes is a thing, and we routinely test for it), it just makes any other health challenge much dicier.

> Diabetes is almost never the direct cause of death, but it’s damn hard in the body. Hurts circulation, hurts your nerves, hurts your organs… and it can flare badly in reaction to some medications. So, the cause of death won’t be “diabetes”, it’ll be heart attack, stroke, chronic infection, something like that.

It's not like the CDC doesn't know this. They understand how diabetes works; that's reflected in the tabulation of data. It's still incorrect to say that the drop in cancer deaths is explainable by the increase in diabetes deaths, not the least because age-adjusted diabetes deaths decreased during most of that interval.

You have to got to decide a parameter at some point. It will never be 100% descriptive in reality, biology is a messy science.

Seems less of a problem, more of a success.

> If our testing were perfect such that we could detect all cancers early enough to stop, that would functionally be a cure, no?

no. being able to detect cancer doesn't mean people are getting tested twice a year for their whole life. for example, you wouldn't have routine cancer screenings for children, but sometimes they get cancer, and on those occasions it'd be nice to have better treatment options.

> it'd be nice to have better treatment options.

No disagreement with you there. By "perfect screening" I assumed a "perfect" patient set where problems brought up at annual checkups lead directly to tests which lead to detection, but perhaps that's a naïve assumption to the point of uselessness.

Still, I think it's easy to write off screening advances as skewing the data set vs. actually saving lives, when they do in fact help people that otherwise could have been much sicker.

We wouldn’t normally even consider talking about these things as if they’re in any way mutually exclusive, would we? Like normally we would all understand that it’s good for cars to get safer and for trauma care to get better.

if cancer screenings for children were painless, lower risk than the cancer, and cost five dollars, you would totally have routine cancer screenings for children

i mean maybe not in the usa but in sane places

They would have them in the US as well, but only for people in the Gold tier of insurance

And only children of parents who were members of whichever political party had decided the scans aren't an evil hoax.

The thing we don't know is which early cancers actually progress to advanced cancer. It's possible that we detect more things that look like cancer, but that would not have ever progressed into full blown metastatic cancer (at least, not before the patient died of something else). So that could artificially make it look like we've gotten better at treating cancer, when actually we've just gotten better at detecting non-life-threatening cancer.

> So that could artificially make it look like we've gotten better at treating cancer, when actually we've just gotten better at detecting non-life-threatening cancer.

No, because the death rate is measured at the population level, not at the incidence level. Fewer people are dying today of cancer than 30 years ago.

The only way that detection bias could impact that outcome is if you make the argument that a large number of terminal cancer cases were going undetected in 1991 at the time of death.

Reply on news.ycombinator.com