Hacker News Reader

Comment by stevenwoo | Hacker News Reader

I don't know much about statistics - can someone tell me why five year survival rates are always used for cancer? This seems kind of off, that instead one should be using the average or median age to which one is expected to live versus how long a cancer survivor lives. Five years seems most misleading when applied to children that get cancer.

View on HN

Traditionally, cancer was considered "cured" when the patient survived 5 years without evidence of disease. We now know that quite a few cancers, including breast, have a significant rate of late recurrence, ie, after the five-year window. But it's still a very strong psychological milestone for the patient and a convenient and practical benchmark for the clinician.

It's my opinion that OS5 is a largely misleading stat that makes us underestimate the role early stages of cancer like ADH/ALH have in recurrence.

That being said, there are a number of reasons that I believe it's so prevalent.

(1) 5 year long survival studies are long and risky affairs, increasing that number just makes it longer and riskier

(2) particularly when speaking about differences in treatment, a statistically significant average increase of 1 month survival over standard of care is sufficient to supplant it, so OS5 v. OS10 doesn't change much if your treating advanced cancers that result in deaths before 5 years.

When it comes to most cancers though, I think OS5 is a very bad stat.

Warning: not an expert. Someone smarter/better educated, correct me (or better, tell me I'm right =)

Cancer tends to (though not always) correlate with age when you break it up by type of cancer. For example, children don't get prostate cancer or lung cancer all that often. Similarly, Retinoblastoma and other types of cancer are more common amongst children. So, because of that, you have a built-in tendency to control for age and allow the 5-year statistic to be more of a valid comparison. Similarly, you can calculate 5 year survival rates for any set of subgroups you want – break it up by age, race, gender, co-morbidity or even cancer variant (many of those are frequently done) and then compare. At some point (usually in pretty quickly) the 5 year survival rate is a meaningful comparison...

Also, to be honest/in my opinion, 5 years is a meaningful number for the lived experience. You can imagine doing a lot in five years vs, say, 6 months – so for that reason, it's a useful metric for generally understanding "what chance do I have at having more life to truly live?"

Tradition. Each study is a riff on past studies. To make things more comparable, you want to change only enough to make your study novel. Therefore, you often use the same analytical methods and report the same indicators as past studies.

5-year survival is just one measure of health. You proposed two others, but none of them are the complete picture. They're all about the same really. Best is actually looking at the distribution of each metric you care about, but then those curves are hard to compare.

5-yrs mortality rates are a questionable statistic, especially when comparing death rates in time. Critics would argue that the lead time (i.e. the time between diagnosis and clinical symptoms) has changed in time due to earlier diagnosis but the effective time of death has not. It's a long winded debate.

Breast cancer mostly affects women 45+ -- or younger women with a genetic disposition.

But otherwise you'd get a lot of age related deaths? If you use a 15 year window, many people with cancer treatment at the age of 70 won't survive the 15 year period. That makes cancers 'deadlier' where they affect older people. A 5 year horizon is probably more suitable to reduce unrelated deaths.

The streetlight effect? If you want to measure median age, you're not going to get the data for a while.

These studies are darn expensive. To track people for 10 years would almost double the cost of it.

Maybe even greater is the opportunity cost. If you have a good result, it would take the double of time for the society to benefit from it.

Because otherwise you couldn’t lobby for early screening programs. Only so few women will get breast cancer and need treatment, while every woman needs to be screened.

When you look at medicine as a business based on billing all the needs justifications make so much more sense.

Agreed. Follow the money.

Also, why has the threshold for high blood pressure moved lower over the last 20 years (multiple times)?

Reply on news.ycombinator.com