News & Views | Published:


Risk factors and random chances

Nature volume 517, pages 563564 (29 January 2015) | Download Citation


The discovery that the estimated number of stem-cell divisions in a tissue correlates with cancer incidence suggests that the varying probability of developing cancer in different tissues is mostly down to random mutations.

Cancer arises through the accumulation of molecular changes that together allow cells to grow in an uncontrolled manner. Many factors contribute to this process, including hereditary genetic mutations and environmental hazards, such as exposure to smoking or radiation. But cancer can still emerge in the absence of these factors, as a result of random mutations that arise during cell division1. Furthermore, little is known about how each risk factor affects different tissues, in which cancers arise at varying frequencies. Writing in Science, Tomasetti and Vogelstein2 show that about 65% of the differences in cancer incidence between tissues can be simply explained by the estimated total number of stem-cell divisions in those tissues. This result suggests that, rather than environmental and heredity influences, stochastic accumulation of mutations during DNA replication is the major cause of variations in cancer incidence between tissues.

Estimating the number of stem-cell divisions that occur in a tissue is a complex task. Tomasetti and Vogelstein performed an extensive literature search to obtain information about stem-cell numbers in various tissues. They then used mathematical and statistical models to estimate the number of stem-cell divisions in each tissue during the average human lifespan, and correlated this with publicly available cancer-incidence data. Although some cancers are certainly driven by genetic predisposition or environmental risk factors, the authors' analysis indicated that these factors explain only one-third of the variation in cancer incidence between tissues (Fig. 1). The rest of the variability was explained by the different number of stem-cell divisions estimated to occur in different tissues. However, some commonly occurring cancers, such as breast and prostate, were not included in the analysis, and these must be examined when data become available. Importantly, although this paper attempts to explain the variation in cancer incidence among tissues, it does not try to quantify the percentage of cancers that arise owing to random accumulation of mutations alone, which is a different measure.

Figure 1: Mechanisms of cancer development.
Figure 1

a, Exposure to environmental carcinogens can cause genetic mutations in a healthy cell population, leading to the generation of cancerous cells (two mutations are shown as sufficient for tumour growth here, although more are often required). b, If a mutation associated with cancer has been passed down from parents to offspring, the offspring has a hereditary predisposition to that cancer. Tumour growth is initiated if any of these cells randomly acquires a second mutation during cell division. c, In the absence of environmental risk factors and genetic predisposition, acquisition of random mutations alone can be sufficient to cause cancer. Tomasetti and Vogelstein2 report that this is the major cause of the variable rates at which cancer arises in different tissues.

This study suggests that processes governing the evolution of cancer cells need to be better understood, so that they can be manipulated to delay the onset of cancer. One drug that might affect evolutionary dynamics is aspirin, which protects against a variety of cancers3. Aspirin modulates evolutionary parameters that determine the rate of cancer development, including cell-division and cell-death rates4, and the persistence of cells with elevated mutation rates5.

In addition to experimental studies, mathematical models are crucial for investigating the evolutionary dynamics of cancer6. Such models have led to the development of several principles that define how random mutations and subsequent selection influence the emergence and growth of tumour cells in certain settings. For example, stochastic evolutionary models have been used to study the emergence and spread of tumour cells in healthy tissue7,8. Furthermore, models that take into account stem-cell dynamics have been used to study the evolutionary pathways by which cancerous cells escape the constraints of tissue regulation9.

Nonetheless, much remains to be done, and Tomasetti and Vogelstein's paper highlights some intriguing directions for future research. For instance, although the authors assumed that random cancer-causing mutations occur during stem-cell division, it is actually unknown in which tissues stem cells (rather than more-differentiated cells) become cancerous10. The correlation they report could arise either way, because the number of stem-cell divisions probably correlates with the number of divisions undertaken by the differentiating daughters of stem cells.

Another gap in our knowledge concerns the relative importance for cancer evolution of embryonic cell divisions compared with divisions after birth11. Although many cell divisions occur after birth in most tissues included in this study, almost no cell division occurs after birth in the cells that give rise to glioblastoma12, an aggressive brain tumour. But Tomasetti and Vogelstein found that the incidence of glioblastoma fits the same trend as the other cancers they studied.

Last but not least, we need to understand how various environmental selection pressures influence the fate of the mutant cells that are generated by chance. Microenvironmental conditions, such as immune responses, inflammation or the presence of cancer-causing molecules, can affect the fitness of specific mutants and thus their ability to grow and give rise to disease. These conditions can change during ageing, resulting in environments that are conducive to the growth of cancerous cells13. Although the correlation observed by Tomasetti and Vogelstein is certainly intriguing, and although accumulation of mutations clearly has a central role in causing variability between tissues, much research is needed to disentangle the complex, multifactorial interactions that result in disease.

A landmark paper14 published in 1981 suggested that most cancers could be averted by removing various lifestyle, behavioural and environmental risk factors prevalent in the population, and thus that much of the risk of cancer could be controlled. Risk factors are certainly involved in promoting the occurrence of many cancers, especially the more common ones. It therefore makes sense to use the presence or absence of risk factors to identify individuals who should be screened for cancer. If, however, random genetic changes are more-relevant drivers of carcinogenesis, then biomarkers and early-detection methods will have to be developed to prevent cancer mortality in the general population.

This might be problematic. Screening for many of the cancers included in Tomasetti and Vogelstein's study is currently difficult, particularly for rare cancers. Screening is successful only when several criteria are fulfilled: that tests are available that detect early disease; that the test's sensitivity and specificity for the given cancer is high; that people are willing to be screened; and that effective treatments exist that can be applied to early-stage cancers to prevent death. Furthermore, the benefits of screening must outweigh the risks of offering screening to the general population. The 'number needed to screen' to prevent one cancer death must be feasible, given the resources required to perform screens on the population as a whole (see

In summary, Tomasetti and Vogelstein's findings emphasize the role of basic evolutionary mechanisms in cancer development, and might lead to new chemoprevention therapies that slow the evolutionary processes at work. Although screening according to risk factors remains a crucial intervention strategy, an improvement in our cancer-prevention efforts in the general population will require the generation of early-detection techniques.



  1. 1.

    & (eds) The Genetic Basis of Human Cancer (McGraw-Hill, 2002).

  2. 2.

    & Science 347, 78–81 (2015).

  3. 3.

    et al. Lancet 377, 31–41 (2011).

  4. 4.

    , , , & Clin. Cancer Res. 9, 383–390 (2003).

  5. 5.

    et al. Proc. Natl Acad. Sci. USA 95, 11301–11306 (1998).

  6. 6.

    & Dynamics of Cancer: Mathematical Foundations of Oncology (World Scientific, 2014).

  7. 7.

    et al. Proc. Natl Acad. Sci. USA 99, 16226–16231 (2002).

  8. 8.

    , & Proc. Natl Acad. Sci. USA 100, 14966–14969 (2003).

  9. 9.

    , & Proc. Natl Acad. Sci. USA 108, 18983–18988 (2011).

  10. 10.

    & Cell Cycle 3, 1558–1565 (2004).

  11. 11.

    & Nature 422, 494 (2003).

  12. 12.

    , , , & Cell 122, 133–143 (2005).

  13. 13.

    , & Aging 6, 1033–1048 (2014).

  14. 14.

    & J. Natl Cancer Inst. 66, 1192–1308 (1981).

Download references

Author information


  1. Dominik Wodarz is in the Department of Ecology and Evolutionary Biology, University of California, Irvine, California 92697, USA.

    • Dominik Wodarz
  2. Ann G. Zauber is in the Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, New York 10065, USA.

    • Ann G. Zauber


  1. Search for Dominik Wodarz in:

  2. Search for Ann G. Zauber in:

Corresponding authors

Correspondence to Dominik Wodarz or Ann G. Zauber.

About this article

Publication history



Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing