Cancer Is More Bad Luck Than Bad Behavior, Study Says | Hacker News Reader

Cancer Is More Bad Luck Than Bad Behavior, Study Says

bloomberg.com

70 pointssvtrent37 comments

This article misstates the finding of the research, as does much of the other coverage. A good explanation can be found here: http://www.theguardian.com/science/grrlscientist/2015/jan/02...

Thank you very much for that helpful link. If you submit it as a new submission to Hacker News, I'm sure I will not be alone in being glad to upvote it to main page visibility.

Thank you. Submitted.

Here's a curve that represents the rate of cancer diagnosis by age in the UK. http://www.cancerresearchuk.org/cancer-info/cancerstats/inci...

If cancer were only about a mutation happening randomly, you'd expect the curve to actually slope downwards, like an exponential distribution.

But maybe cancer requires several mutations to happen in a row, which would suggest an Erlang distribution of which we only see the beginning of the hump. 20 mutations give a very good curve fit (1).

An alternative hypothesis is senescence. Maybe the mutation rate itself increases with age, or maybe the immune system isn't as able to nip cancerous growths in the bud.

This is important! Whenever the impact of a disease doesn't follow a chi squared or exponential distribution, it suggests two ways to fight the disease. Directly target the disease itself, or target the process which over time makes us more susceptible to it.

(1) warning, back of the envelope, there are plenty of other effects to account

That looks like a perfectly fine exponential distribution.

No it doesn't. The exponential distribution p.d.f. is decreasing. http://en.wikipedia.org/wiki/Exponential_distribution#mediav...

But that's not what we are interested in, no? Nobody asks "whats the probability that you'll get cancer at precisely age x", they ask "what's the probabiblity that you'll have cancer by age x". I mean, we often can't even determine at what precise age the cancer developed. That's a problem for the PDF, not so much the CDF.

The graph I linked to plots the PDF. It's not exponential. It indicates something is getting worse over time.

We know quite clearly that cancer is (generally) caused by a set of mutations - not necessarily in an order, but some orders are not successful. There are four or five genes which keep social order amongst the other genes. If you silence all of these, you get cancer.

Cells that deviate from what they're supposed to do are either (in order), repaired, silenced, voluntarily commit suicide, or are killed. There are proteins (genes) that are the final judges for each of these processes - and have 'go, no-go' power. Only if all of these judges are killed do you get a cell that can do anything it wants - like replicate uncontrollably to the detriment of the host ('cancer'). Thus the statistics of getting cancer roughly follow the idea that you have to get random DNA modifications of those exact 5 genes, in a single cell. Lots of things can increase your random modification rate (UV, smoke, radiation, etc). Some of these things correlate though - and again, what hurts one cell, might hurt its neighbor just as bad. They're not entirely independent events. For example, losing your DNA repair machinery (this is what HPV does - it silences your DNA repair machinery) amps up the baseline mutation rate and makes further mutations more likely (dependent correlations then arise).

The Brca gene that has caused so much controversy in patent law (whether a test for its existence could be patented) and indicates whether a person might or might be susceptible to breast cancer, is the master repair technician of the cell. In people who have this gene in working order, the Brca gene signs off on whether the cell is in need of repair. But if the Brca is not it working order, cells that are in need of repair might not get it, and instead are allowed to more freely operate under non-optimal internal conditions. If you are missing or have a mutated version of Brca, you are missing one of the checkpoint processes.

So again, we quite clearly know of a handful of genes which do most of the master regulation of a cell's job - and if these jobs go unfulfilled - by having their blueprints be damaged by the environment - you have fewer and fewer mechanisms to prevent that single cell from runaway growth.

I see the sole previous top-level comment here [edit: which has now been deleted] started out with a misstatement of fact. In fact, age-adjusted cancer death rates have been steadily dropping in the United States[1] and in general death rates from cancer and from other major causes of death have been decreasing throughout the developed world throughout my lifetime.[2] What that means in practical terms is that life expectancy at age 40, at age 60, and at even higher ages is still rising throughout the developed countries of the world, so if you have reached the age of 50, you can be ever more confident of reaching the age of 65, and if you reach the age of 65, you can be ever more confident of reaching the age of 80, and so on.

It is possible for expanded cancer screening programs to produce the illusion that rates of age-adjusted cancer diagnosis have increased because of a real increase in cases of cancer,[3] but what is really going on with increased screening is increased detection of early cases of cancerous growths that eventually die off and never cause clinical disease. The human body has some defenses against cancer, and many cancers reach a growth dead-end not long after starting. Screening programs sometimes detect harmless cancers that don't need any treatment except watchful waiting.

[1] http://www.reuters.com/article/2014/12/31/us-cancer-casualti...

http://www.bloomberg.com/dataview/2014-04-17/how-americans-d...

http://www.slate.com/articles/health_and_science/science_of_...

[2] http://www.nature.com/scientificamerican/journal/v307/n3/box...

[3] http://www.sciencebasedmedicine.org/recent-developments-and-...

Different cancers are different. Skin cancer, both basal cell and melanoma, is usually caused by the sun. Other cancers have a very strong genetic correlation. This study did not consider the most common cancer in men, prostate cancer, or breast cancer. It's interesting that random cell division is such a strong factor in cancer, but it is completely incorrect to say that luck has a bigger effect than behavior for all types of cancer.

This is an excellent comment, but I want to point out that getting a cancer that has a strong genetic correlation is still "luck" rather than "behavior" from the perspective of a cancer patient.

There is a major difference, which is that genetic correlations can be tested for, and a strong one can indicate that frequent testing is needed, catching cancers earlier when they're more treatable. For many cancers, catching them early is critical.

The tricky thing with correlation is that behavior can still act as a "middle-man" for genetic correlation. For instance, a genetic susceptibility to addictive behavior could lead to higher rates of smoking, and therefore lung cancer. (While some of the more obvious factors might be corrected for, there's no way that all of them would be.)

I once heard stress has a big role in cell division quality. Could it be they merely didn't take stress into account and credited luck instead?

There are a variety of basic errors in the interpretation of this study:

http://ameyer.me/science/2015/01/02/vogel.html

The tl;dr; version: this 65% number comes from a correlation coefficient on a log-log plot, which means the correlation coefficient has nothing to do with the actual fraction of cancer cases one can attribute to stem cell division numbers.

This is interesting, but dangerous. Smokers make up 90% of lung cancer deaths [1]. At some point you have to acknowledge that smoking definitely does something to influence lung cancer.

[1] http://www.lung.org/stop-smoking/about-smoking/health-effect...

Also, exercise reduces colon cancer risk by 40-50% and breast cancer risk by 30-40%. Then there are other risk factors such as poor diet, insufficient vitamin D, etc. Overall it seems highly unlikely that cancer is mostly "bad luck".

http://www.cancer.net/navigating-cancer-care/prevention-and-...

50% of what though? If you have 1 in a million chance, increasing that to 1.5 in a million isn't a big change. If "bad luck" afflicted 1 in 100,000 then then bad luck would still have a much bigger effect, despite lifestyle changes clearly being significant.

(These numbers totally made up)

Do these percentages take into account causation vs correlation? Apparently (surprisingly), it's still very difficult to prove causation without a laboratory experiment.

Perhaps people who are likely to exercise have a certain genetic disposition to begin with.

The Bloomberg article mis-interprets the scientific article.

In effect, the scientific article shows how we could predict which tissues are most likely to become cancerous if we had no population statistics on cancer. That is, we could look at the rate of cell divisions in a tissue and say "Oh yeah, divisions occur frequently here, so there are going to be more cancers in this particular type of tissue."

The scientific article does NOT say that the cancer which has arisen in random person X is mostly due to "bad luck" in person X. If person X smokes 4 packs of cigarettes per day, the person's risk of lung cancer is certainly affected. Of course, there is still some element of luck (not every heavy smoker gets lung cancer), but you can think of it like using loaded dice at the time of each DNA replication... loaded dice don't turn up snake eyes every time, but they certainly change the odds, and if you were playing craps there is no way you would want to use loaded dice.

My reaction to the scientific article is... yawn. The result could have been predicted. Cancer is a disease of DNA replication. It makes sense that clinical problems arising from DNA replication would occur most often in places having the most DNA replication events. But it's nice to have experimental verification.

We pin-pointed DNA double-strand breaks using sequencing, and found very high correlation with genes damaged by cancer. For e.g. the "ERG" gene which fuses with TMPRSS2 in prostate cancer was one of the genes in our list. http://www.nature.com/nmeth/journal/v10/n4/carousel/nmeth.24...

More research should reveal that it's more than just pure luck :)

Immune system kills spontaneous blood cancer cells every day - http://www.medicalnewstoday.com/articles/272092.php

Mutations do happen during replication. But is it random? How about underlying mechanisms involving telomeres or histones? Or something else.

That's just what they want you to believe. They being Corporate America who profit enormously when you eat their processed foods, smoke their cigarettes, and take their drugs. Cancer is mainly caused by environmental toxins, so if you don't want to get cancer you have to detox: eat only raw paleo or vegan, cleanse with lemon juice or unsweetened tea four times a week, and for God's sakes man, DON'T BREATHE THE AIR!!!

Reply on news.ycombinator.com