Publishing in Science a group of researchers studied the relationship between the number of normal stem cell divisions and the risk of 17 cancer types in 69 countries. They reported a strong correlation between cancer incidence and normal stem cell divisions in all countries, regardless of their environment.
Prof. Mel Greaves, Director of the Centre for Evolution and Cancer, The Institute of Cancer Research, London, said:
“The significant correlation between stem cell turnover in tissues and risk of cancer (worldwide) is to be expected, whatever the mutational mechanisms. More stem cell proliferation provides more target cells for mutational change and the development of cancer. This is why young children and adults have very different cancer types originating in different cells and tissues.
“As in their previous paper in Science the authors introduce some significant errors and omissions. Prostate cancer has a significant component of inherited risk (more than most other cancers). The authors only consider mutations (SNVs) due to replication errors despite many of the critical driver mutations in cancer being gene copy number variations or chimaeric fusion genes. These are caused by DNA strand breaks and can have internal or external causes.
“Even if, as this study suggests, most individual cancer mutations are due to random chance, the researchers admit that the cancers they cause may still be preventable. We have good evidence to show that cancer is caused by a complex mix of environmental exposures, inherited risk, and random chance. And while the genes we inherit from our parents are unreturnable and many chance events are non-negotiable, fortunately for us, exposures are major contributors to our risk of cancer and offer a route to risk reduction or prevention.
“Even if the authors are correct in their estimates, the relevance for cancer prevention here is obscure. Cancer Research UK estimated some years ago that just over 40 per cent of cancers had external causes that were both identifiable today and, in principle, avoidable. That estimate excluded several of the most common cancers. What should matter at least as much for public health is the fraction of total cancer deaths that could potentially be preventable and the estimates for this are 75 to 95 per cent. We therefore need more insight into causal mechanisms in breast, prostate and colorectal cancer for example. It’s important that findings like this don’t cause us to lose sight of the important public health messages around cancer prevention and the need for continued research to learn how we might further reduce our risk.”
Prof. Lawrence Young, Director of Cancer Research Centre, University of Warwick, said:
“Can an understanding of the sources of gene mutations help us to explain cancer risk and to develop better approaches to cancer prevention? This paper contributes to the controversial debate about the relative impact of extrinsic (environmental) versus intrinsic (heredity, DNA replication errors) factors in causing the accumulated gene mutations that drive cancer development. It concludes that gene mutations due to random DNA replication errors are responsible for nearly two-thirds of the mutations found in cancer cells meaning that not all cancers can be prevented by avoiding environmental risk factors. The multifactorial nature of cancer makes reducing risk analysis to a simple set of mutation drivers difficult. Any model attempting to fully describe cancer risk needs to take into account a variety of extrinsic factors that influence cancer development and progression without inducing gene mutations such as inflammation, immune responses and the tumour microenvironment. So while this study is useful in attempting to integrate epidemiological and genome sequencing data, the message is complex and does not diminish to need to focus on improved approaches to both primary and secondary cancer prevention.”
Prof. Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“I don’t think anyone now would seriously disagree that there are three possible causes of the mutations that lead to cancers – environmental factors like tobacco smoke, heredity (“it’s in your genes”), and random errors that arise when DNA is replicated in the body as cells divide. The issue is the relative importance of these three.
“There’s a back story here. At the start of 2015, two of the authors of this new study, Cristiano Tomasetti and Bert Vogelstein, published a paper in Science that was heavily reported and, I’d say, widely misunderstood. That study estimated that 65% of the variability in cancer risk between different body tissues that could be explained by the pattern of divisions in the cells that make up the tissues. The 65% was subject to a lot of statistical uncertainty, but the main problem was that people reported this result as meaning that 65% of cancers were caused by random replication errors. It doesn’t mean that at all (though it does have a respectable, if slightly obscure, statistical meaning). The challenge of estimating what percentage of cancer cases were due to different categories of cause was taken up by others, notably in a Nature paper by Wu, Power, Zhu and Hannun, that came out in January 2016. Wu and colleagues didn’t deny that random replication errors had any role, but they re-asserted that a large proportion of individual cancer cases are due to the effects of environment. Since then, others have pointed out some flaws in this Nature paper and continued to investigate.
“This new paper by Tomasetti and colleagues brings in data from across the world, rather than (as in their previous paper) just from the USA. What I most like about it is that it provides some much-needed clarity, in making a clear distinction between the roles of environment, heredity and replication errors in producing the mutations that lead to cancer, and the extent to which different types of cancer might be prevented by removing environmental and other non-random causes. You can’t really allocate just one cause to a specific cancer case, because you don’t know where the necessary mutations came from in an individual case. They might have come from any combination of environment, heredity and random replication errors. On a statistical basis, you can answer questions like “what proportion of cancers would not happen if we could make (for example) environmental causes disappear”. Really the key point of the paper is that (according to the authors) it is true that a large proportion of the mutations that can give rise to cancer are random replication errors, and so (in a sense) “just bad luck”, but nevertheless you could still remove a large proportion of cancers by removing the environmental causes. For instance, they estimate that somewhere around a third of all the mutations that might lead to a particular kind of lung cancer are the result of random errors in cell replication, but point out clearly that this doesn’t change the fact that about 90% of lung cancers of this type could be prevented by changing the environment (principally, by removing tobacco smoke). In other cancers, the percentages are very different, but the key point is that, for a given type of cancer, the percentage of preventable cases can be high even though a substantial proportion of the mutations that potentially lead to the cases cannot be prevented because they are caused by random cell replication errors.
“This sounds rather odd, but the explanation is that it takes several mutations for a cancer to arise and progress. In the model described in the paper, they assume it takes three mutations. If there were no environmental or hereditary causes of mutations, then many people who have just one or two mutations from random replication errors, or none, and who currently might get the necessary “three strikes” to get a cancer by having some more mutations from environmental or hereditary causes, won’t get cancer because those hereditary or environmental causes would no longer be there. So despite the role of the random replication component in producing mutations, you could still reduce the cancer risk hugely (for many types of cancer) by getting rid of the environmental and/or hereditary causes.
“This paper certainly won’t be the last word on these matters. The accompanying “Perspectives” article by Nowak and Waclaw makes it clear that, while we certainly need mathematical modelling to understand what’s going on, the models that have been used so far have some important shortcomings. Many of the details of the modelling and data in the new Tomasetti et al. paper are in the supplementary materials, not the main paper, and I have not yet been able to consider those. It’s likely that some of the details of their results will be challenged, and in any case some of the estimates are still subject to considerable statistical uncertainty. One awkward issue is that the different potential causes can be correlated with one another – if you change the environment, for example, the rate of random replication errors could conceivable change too. But one key point will surely remain: when talking about percentages due to different factors in cancer causation, you have to be extremely clear about exactly which percentage you are actually talking about.”
* ‘Stem cell division, somatic mutations, cancer etiology, and cancer prevention’ by Tomasetti et al. will be published in Science on Thursday 23 March.
Prof Mel Greaves: No conflicts of interest to declare.
Prof Lawrence Young: No conflicts of interest.
Prof Kevin McConway: No conflicts of interest.