Most types of cancer not due to “bad luck” [pdf]

On the political front, the document encourages people to reject that cancer predilection is stochastic and to encourage the continued study of cancer.

The scientific component attempts to provide an underpinning to the political goal. It uses a handful of epidemiological examples, such as the climbing incidence of colorectal cancer in Japan over a genetically-insignificant timescale (which suggests an environmental factor at play).

The article fails to support its headline ('Most types of cancer not due to "bad luck"') through any scientific instrument. While its policy goal (the continued pursuit of the causes of cancer) is reasonable, nobody is saying that we now know that cancer is predominantly caused by bad luck so we should stop studying it (certainly not Vogelstein, whose career is built on the study of cancer).

Overall, it's not clear to me why this was published.

The original paper (here: http://www.sciencemag.org/content/347/6217/78.abstract -- the abstract is enough to deduce the conclusion) is a simple statistical analysis of mutation probability in certain cell types. It's basically saying that many cancers are not necessarily influenced by external factors and can be predicted by nothing more than the number of cell divisions and the probability of DNA mutation compounded.

That, in itself, is a fascinating scientific result. Arguing against the statistics here is probably not a good idea, unless there was a serious error in the math or an order of magnitude error in the probability of mutation. Certainly there are other possibilities of error in the data or methodology.

But the article in opposition (OP) is doing something odd that doesn't appear to be a scientific argument: it's railing against the language and the implications of the very concept of statistical analysis applied to the problem. It's assuming that because we discovered that, perhaps, cancer has a certain probability of occurring independent of external factors, that it will somehow slow research or cause us to throw up our hands and give up on prevention strategies or research.

I don't think that's the case. Rather than understanding the absolutely fascinating statistical analysis here, the OP article comes across as reactionary and unscientific.

Assuming the data is sound, the statistical analysis is profound. Most often, statistics such as that are profoundly misunderstood as well: people have an incredible capacity for attribution bias and data disbelief. This simply underscores the need for a better education in basic statistics as well as science across the board.

E.g.: You're a smoker who develops lung cancer: you are 'unlucky' in the sense you suffered a random mutation AND also 'responsible' because you smoke, which is a primary risk factor.

Most 'types' of cancer are unequivocally about bad luck; nearly any cell line in your body can develop overproliferative mutations. There are hundreds of 'types' of cancer.

'Cases' and deaths from cancer are, globally and in the US, more environmentally influenced, because of the huge impact of smoking on lung and colorectal cancer incidence and mortality in particular.

This is just my opinion, but I think all of the following probably help: try to have much less stress in your life; plenty of exercise; prefer organic food if available; avoid deep fried food; go light on barbecuing, as tasty as it is; eat lots of and a wide variety of green vegetables.

Our analysis shows that stochastic effects associated with DNA replication contribute in a substantial way to human cancer incidence in the United States.

First, the terminology is important to understand. In medicine, "stochastic" is in contrast to "deterministic" which describe the occurences of pathology based on a risk factor. In a "stochastic" model, if you receive 10 severe sun burns then your chances of getting a skin cancer go up over someone who only received 1 severe sun burn, but are by no means certain or even that your cancer will be more severe... simply an increase in probability. In a "deterministic" model, if you receive 5 Grays of radiation then you will have less severe symptoms than someone who received 50 Grays of radiation and for the most part, your symptoms will be certain and predictable.

Now you need to have one element of cancer biology under your belt:

http://en.wikipedia.org/wiki/Carcinogenesis

This basically says that as cells divide, errors are propagated through their progeny until there is a proverbial 'straw that breaks the camel's back' and the cell becomes cancerous.

In summary, that line from the final paragraph is basically saying that cells that divide more rapidly in the human body are more likely to become cancer. Well, yeah of course!

We have known this for a long time but what this paper did is sort cancers into two groups, one that is primarily stochastic and one that is primarily deterministic. The result appears impressive because they have a lot of cancers on the stochastic side... stating that their incidence is more predicted by cell division rather than the deterministic, preventable factors.

But as the WHO correctly stated, they are incorrectly analyzing this data as described in this paragraph:

These include the emphasis on very rare cancers (e.g. osteosarcoma, medulloblastoma) that together make only a small contribution to the total cancer burden. The report also excludes, because of the lack of data, common cancers for which incidence differs substantially between populations and over time. The latter category includes some of the most frequent cancers worldwide, for example those of the stomach, cervix, and breast, each known to be associated with infections or lifestyle and environmental factors.

The cancers grouped into the stochastic side overall make up a tiny percentage of all cancers and so by number/%/incidence, the deterministic side would be enormous if graphed appropriately. For example, their most 'stochastic' cancer is pancreatic islet cell which only has 2500 cases/year in the US vs their 'lung smokers' cancer which has >220,000 cases/year.

I mean they seem almost willfully blind to their incorrect conclusion. Environmental factors are HUGE and the cancers caused independent of those would not benefit from screening and environmental risk reduction--a fact we already know!

As a last note, the 'most' deterministic cancers--inherited types--have to be agressively treated with screening, prophylactic measures and environmental exposure risk reduction, etc. You can read about a couple of the deterministic cancers they mention in this paper:

http://en.wikipedia.org/wiki/Familial_adenomatous_polyposis

http://en.wikipedia.org/wiki/Lynch_syndrome

[0] http://www.nytimes.com/2012/10/28/magazine/the-island-where-...

> In principle ... nearly half of all cancer cases worldwide can be prevented

I am not sure what to take from this article. Lung cancer prevention is already in progress. At the same time prostate cancer has even higher mortality (for non smokers) and preventive checks are not even covered by most insurance companies.

The paper also technically sets out to explain why cancer incidence varies between different tissues. An analogy is correlating the average temperature with distance from the equator. It is clear that most of the difference in the mean temperature of Equador compared to Iceland is explained by distance from the equator. You can create categories like number of days with rain, number of days with a temperature > 30 degrees C, and each of these values will be highly correlated with distance from the equator. But it does not follow that this strong correlation means that there are hardly any environmental influences on the temperature in St Petersburg tomorrow, and that we can fire all the meteorologists.

(Disclaimer for what follows: I do life sciences research, so I may have an overly pristine view about physics.)

The really interesting thing about all this I think is the collision of maths/physics/engineering and life sciences! Vogelstein wrote this paper with a mathematician. They basically did a back of the envelope calculation, an approach that is much favoured by engineers but completely alien to life science researchers. These simple calculations are useful because in the physics/engineering paradigm, abstractions are incredibly powerful and non-local. For example, many important physical laws that apply equally across many orders of magnitude, can be derived from thinking about falling apples or billiard balls. Paul Dirac predicted the positron by simply exploring other possible solutions to an equation! To me, this is an extraordinary and winning moment for quantum physics, where the theory is so powerful, that it can drag us screaming towards completely unintuitive and otherwise inaccessible conclusions. To a biologist, this kind of thing is ridiculous and alien. The same abstractive power is almost completely absent, and both experimental and theoretical models are extremely limited.

There are many reasons for this. You could argue that biology is not amenable to the often time-invariant abstractions so useful in the other sciences. Or perhaps biologists aren't trained to think that way. In any case, it is interesting to watch when an engineer encounters biology, and this paper is an example. I am still not sure whether we need more engineers in biology or not?

I think it is abundantly clear though that breakthroughs in biology don't come from professors having epiphanies while walking amongst the pine cones on a cloudy autumn day. Human intuition alone is failing to get us very far in biology. We need a paradigm that can deal efficiently with uncertainty, incomplete yet massive data, noise, simulation of highly parallel processes on long time scales, causality in networks of highly correlated actors etc - I hope that some future melding of computer science, biology and staggering computational resources will give us more useful ways to investigate life.

But one tenant of scientific thought is that truth and factuality are not measures of an individual's or society's belief, but rather, they are discovered by experiments whose results can be replicated.

> People distrust science when it conflicts with their valued beliefs, or when science suggests a solution or intervention that conflicts with their beliefs. People are happy to trust science when it does not conflict with their ideology or narrative.

I agree, but those same people generally rationalize those views to themselves in the way Adams noted: by pointing at reasons to distrust science or scientists. They get caught up in issues like "climategate", they assert that results about medical science are distorted by financial pressures, etc.

In short, the rationalizations of individuals who reject scientific results often point to the scientists behind the results, their interests and motivations, instead of the results themselves. This isn't exclusively true, but I think the tie in to the idea of "scientific consensus" is this:

Perhaps science would be better served by a decreased emphasis on the individuals carrying out scientific research and increased emphasis on the results themselves.