Does cancer play dice with us?

In 2015, Tomasetti and Vogelstein published in Science a famous and controversial paper on cancer etiology1. They showed that lifetime risk of many different cancers highly correlates with the total number of stem cell divisions of the tissue from where the cancer arises. The correlation between these two parameters was R=0.81.To understand how strong this correlation is, one can imagine that R of value 1 would mean a perfect correlation: one could exactly calculate the value of one parameter just knowing the value of the other. Tomasetti and Vogelstein interpreted these data as a proof of the existence of another player in cancer etiology, along with hereditary and environmental components: intrinsic replication errors. However, they went further. They concluded that a linear correlation of R=0.81 points to the possibility that around 65% of the variations in cancer risk amongst different tissues can be just explained by the number of stem cell divisions in those tissues. In simpler words, two third of cancer-driving mutations are due to “bad luck”.
This paper sparked enormous controversy and many response letters have been sent to the journal Science by experts in the field to criticize Tomasetti and Vogelstein’s claims2,3,4,5,6,7.Sentence 1

Two main criticisms to this work were the following:

  • The analysis was restricted to the United States population, which does not represent the entire human population.
  • The two most frequent cancers, prostate and breast, were excluded from the analysis.

In a new paper published again in Science, the authors faced these two problems drawing the same conclusions of the first work8. Indeed, analyzing the risk of 17 different cancer types (including prostate and breast) in 69 countries, a similar strong correlation (R=0.80) was found. Again, two third of the mutations originates from unavoidable intrinsic mistakes. Nevertheless, authors underlie that “a cancer that required two mutations is still preventable if one of the mutations was due to R (replication errors) and the other due to an avoidable environmental factor”. This concept is extremely important since conveying to the public the misleading information that cancer onset is mainly due to “bad luck” and not to wrong lifestyles, can generate dangerous interpretations, which in turn highly impact the efficacy of cancer prevention strategies and programs.

Tomasetti and Vogelstein work is important to raise awareness regarding a so far underestimated player in cancer etiology: the chance. Humans always strive to find an explanation for everything bad that happens ignoring the power of randomness. Nevertheless, the two above-mentioned works have many flaws, which need to be clearly conveyed to the public in order to avoid unwanted fallouts.Sentence 2

  • These findings are hypothesis-driven and not evidence-driven. In other words, the positive correlation does not tell us anything regarding the underlying biological processes9.
  • There are few and not completely reliable data regarding the dynamics and hierarchy of human stem cells in the different tissues. Tomasetti and Vogelstein pool mouse and human data for some tumors, without mentioning the existence of major species-specific differences (number of stem cell in the tissue, transcriptional regulation, etc)10.
  • Correlation cannot differentiate risks. In particular, the positive correlation between cancer risk and stem cell divisions in a tissue cannot distinguish contributions from replication, hereditary of environmental factors. This is well explained by Wu and colleagues with an easy example. If a mutagen that increases the risk of all cancers is introduced in the environment, the correlation found by Tomasetti and Vogelstein would not change, however, the fraction of environment-induced cancer would substantially increase10.
  • In both studies the authors state that for 22 tumors entities analyzed “primary prevention measures are not likely to be very effective”. A strong statement, not corroborated by strong evidence. Moreover, statistical problems related to the correlative analysis have been raised further pinpointing the lack of certainty on the matter11.
  • These works do not prove causal relation: many pieces of evidence suggests that stem cell division rates, and the related replicative errors, are not simply due to aging and chance. Indeed, many external factors, including diet, obesity, pollution, infections, and inflammation can significantly affect them5.
  • Most cancers show distinctive population profiles, being more frequent in certain zone of the worlds compared to others. Rates of incidence may vary over time, and more importantly, migrants soon show incidence rates similar of those belonging to the host country. This strongly points to the major role played by environment and lifestyles in cancer pathogenesis5,10.
  • Epidemiologic data show that many cancer have a substantial environmental component:
    – 75% of colorectal cancer risk can be attributable to diet
    – sun exposure ascribes to around 65-56% of melanomas
    – around 75% of esophageal or head and neck cancers seems to be caused by alcohol and tobacco.
    – 90% of cervix cancer may be due to papilloma virus
    – Helicobacter pylori may account for 65-80% of gastric cancer.
    These examples are just the tip of the iceberg, and underline the importance of extrinsic factor in cancer onset and the pivotal role played by cancer prevention10.
  • The last and the most important fact (in our opinion) is pinpointed by Ashford and colleagues in a response letter to the journal Science:
    “The fact that age-adjusted cancer rates* for different tissues vary substantially among countries where statistics are kept, and between workplaces or communities that differ in environmental exposures, demonstrates that a large fraction of cancers are influenced by environmental factors”2,12.

The last point is of particular importance as Tomasetti and Vogelstein, while trying to answer many of the above-mentioned criticisms, did not comment on that issue neither in the response letter nor in their new paper13.

At the moment, it is clear that an exhaustive theory of cancer initiation should include other factors yet to be discovered and many questions are out there waiting to be answered:

  • How many genes need to be mutated in a certain tissue to initiate cancer?
  • Can all human cell types repair DNA damage with the same efficiency?
  • Does immune surveillance** play a role? If yes, to which extent and similarly in all tissues?
  • Can systemic conditions, such as stress and depression, influence the overall cancer risk?
  • Should environmental factors be added as single agents case by case in the correlation or as a unique family?
  • Are there any other cancer theories, apart from somatic mutation theory, that can be used to model cancer initiation in a different way?
  • Can mathematical and statistical models provide a reliable framework of a disease whose pathogenesis in human is still relatively poorly understood?

Nowadays, scientific discoveries can have an immediate impact as news can spread at light speed all over the globe. This is crucial to make the scientific community and the public aware of new discoveries and to readily apply them. However, scientific discoveries conveyed by the media to the public can be often misleading, especially when the article’s title proposes statements that over-interpret the actual findings of a study. Sending the message to the public that two thirds of all cancers are due to mere bad luck can be extremely dangerous. This is even more worrisome if we think that in the “social media era”, many people tend to read only the title’s statements without examining the topic in depth.Sentence 3

Scientific discoveries can be a double-edged sword if not carefully handled. It would be advisable that in the near future scientific commissions monitor media articles to avoid the uncontrolled spread of misinterpreted scientific articles to the public. A scientific-review process may represent a powerful vaccine against those articles written with the sole scope of selling copies, and the only way to righteously inform people about Science.

Finally, we want to conclude with a provocative/funny fact.

Chocolate consumption correlates with Nobel prize laureates – Messerli 2012, New England Journal of Medicine

In 2012, Franz Messerli ran a correlative study between chocolate consumption and cognitive functions. The study, published in New England Journal of Medicine, found out that chocolate consumption per capita positively correlates with number of Nobel laureates per 10 million persons, and the R-value is a striking 0.79! If Sweden, an outlier, is taken out, the R raises up to 0.86. Ready to eat some chocolate?




Davide Mangani



* age-adjusted cancer rates: In epidemiology and demography, age adjustment, also called age standardization, is a technique used to allow populations to be compared when the age profiles of the populations are quite different. Source and more info:

** immune surveillance: In recent years, substantial work have demonstrated that immune system can recognize and destroy nascent tumor cells. The existence of this mechanism has been further corroborated by the recent success of novel therapeutic strategies employing immunotherapeutic means to fight cancer. More info:


1: Tomasetti and Vogelstein 2015, “Variation in cancer risk among tissues can be explained by the number of stem cell divisions”. Science.

2: Ashford et al 2015, “Cancer risk: role of environment”. Science.

3: Wild et al 2015, “Cancer risk: role of chance overstated”. Science.

4: Potter and Prentice 2015, “Cancer risk: tumors excluded”. Science.

5: Gotay et al 2015, “Cancer risk: prevention is crucial”. Science.

6: Song and Giovannucci, “Cancer risk: many factors contribute”. Science.

7: O’Callaghan 2015 “Cancer risk: accuracy of literature”. Science.

8: Tomasetti, Li and Vogelstein 2017, “Stem cell divisions, somatic mutations, cancer etiology, and cancer prevention”. Science.

9: Nowak and Waclaw 2017, “Genes, environment and “bad luck””. Science.

10: Wu et al 2016, “Substantial contribution of extrinsic risk factors to cancer development”. Nature.

11: Altenberg 2015, “Statistical problems in a paper on variation in cancer risk among tissues”. Preprint at

12: IARC, Cancer Incidence in Five Continents, Vol. X (International Agency for Research on Cancer, Lyon, France, 2013).

13: Tomasetti and Vogelstein 2015, “Cancer risk: Role of environment—Response”. Science.