In graduate school, a wise colleague once told me that, if you want to figure out if a paper is trustworthy or not, don't look at the data—look at the methods. With that in mind, here's an example of how the Tomasetti paper estimates the fraction of mutations resulting from replicative errors:
There is a lot more complicated math involved in other cases, but this description is much shorter, easier to understand, and illustrates the flaws and limitations of the study. The study relies on existing epidemiological data to craft their estimates of the fraction of mutations due to replication error. As one can see above, most of their "calculations" are back-of-the-envelope estimates based on those data. Thus, the starting point for most of their calculations is the premise that ~40% of cancers can be prevented by lifestyle changes. Next, they assume that the other 60% of cancers cases must all be unpreventable, essentially making the assumption that that all causes of cancer are currently known and accounted for. Finally, they make the assumption that all cancer cases not from known environmental or known genetic sources are due solely to replicative error. From these starting points and assumptions, they estimate that 66% of all cancer-promoting mutations are due to replicative error.
Thus, a better way to phrase the conclusion of the paper would be: if ~60% of cancers cannot be prevented, then ~70% of cancer-promoting mutations are due to replicative error (the fraction of unpreventable mutations will always be higher than the fraction unpreventable cancers because most cancers require more than one mutation to arise). The article does not really add much new data to understand the fraction of preventable cancers, but rather just takes already published estimates, plugs those into a couple of equations, and comes out with a rough estimate of the fraction of mutations from replicative error. Therefore, I would echo the sentiment from Martin Nowak I posted previously, and say that the numbers from this paper are very rough estimates that are nowhere close to being set in stone. For example, the
http://www.nature.com/nature/journal/v529/n7584/full/nature16166.html paper from
Nature used a very different method for estimating the fraction of mutations due to replicative error (based on studying the pattern of mutations from DNA sequencing data) and came up with very different estimates. The study also does not rule out the possibility that other factors contributing to cancer could be recognized and increase the fraction of preventable cancers.
It's also worth noting that some of the actual data collected by the authors of the study suggests the fraction of preventable cancers might be somewhat higher. The authors collected data on cancer incidence rates in many different countries and found that cancer incidence in a particular organ correlates well with number stem cell divisions in that particular organ. However,
other researchers have looked at that data and noted that the differences in cancer rates among the countries supports a high contribution of supposed environmental factors (this argument was also made in the Wu et al. paper referenced above). This paper will certainly not be the last we hear of this debate.