Download Are Most Cancers Caused by Specific Risk

JNCI J Natl Cancer Inst (2016) 108(3): djv343
doi:10.1093/jnci/djv343
First published online November 9, 2015
Commentary
commentary
Are Most Cancers Caused by Specific Risk Factors
Acting on Tissues With High Underlying Stem Cell
Divisions?
commentary
Edward L. Giovannucci
Correspondence to: Edward Giovannucci, MD, ScD, Harvard T. H. Chan School of Public Health, 665 Huntington Avenue, Boston, MA 02135 (e-mail: egiovann@hsph.
harvard.edu).
Abstract
A recent paper by Tomasetti and Vogelstein demonstrated a high correlation coefficient of 0.81 between estimated lifetime
normal renewing cell (stem cell) divisions among tissues in the body and the lifetime cancer risk in that organ. This finding
has been interpreted frequently to suggest that if two-thirds of cancers arise primarily through normal proliferation then
environmental and hereditary factors combined could explain only one-third of cancers. Yet, the pool of dividing stem cells
is what risk factors act upon; it is unlikely that risk factors and proliferation act completely independently and are simply
additive; thus, there is no constraint that stem cell proliferation and environmental/genetic attributable risk sum to 100%.
The cancers illustrated to represent lifetime risk in the paper by Tomasetti and Vogelstein all implicitly incorporate risk
factors common in the United States (example, obesity, physical inactivity, tobacco, alcohol, diet, infectious agents). In fact,
there is little evidence that a cancer would exceed a substantial rate, such as greater than 1% lifetime risk, in the absence of
an important risk factor. Relatively high rates of cancer (eg, > 1% lifetime risk) only seem to occur in organs when strong risk
factors (example, 10- to 20-fold) are superimposed on relatively high stem cell division. In organs with low stem cell divisions,
the lifetime cancer risk will typically be very low. The major types and most abundant cancers in a given population will arise
from tissues that have relatively high stem cell division rates and that have a high prevalence of strong relevant risk factors.
A recent paper by Tomasetti and Vogelstein in Science (1) demonstrated a high correlation coefficient of 0.81 between estimated
lifetime normal renewing cell (stem cell) divisions among tissues
in the body and the lifetime cancer risk in that organ using United
States rates. This finding offers important insights into the role
of chance, likely from replication error mutations during normal
cell division in cancer etiology. The authors argue that normal
stem cell division variation across tissues is likely the single most
important factor explaining variation in cancer susceptibility
across tissue types. The authors further interpret their findings
to inform on the preventability of cancers, essentially arguing
that if two-thirds of cancers arise through normal proliferation,
termed “bad luck” (1), then environmental and hereditary factors
combined could explain only one-third of cancers. This specific
interpretation has generated much spirited debate concerning
the preventability of cancer. Better understanding of the insights
offered from this paper requires an appreciation of three main
points that are summarized in this Commentary. The accuracy of
the assumptions made in estimating stem cell proliferation in tissues may be questioned, but this Commentary will only address
the interpretation on preventability under the assumption that
a high correlation between tissue stem cell division and cancer
susceptibility in an organ exists.
Some of the Data Points in the Plot Are Based
on Average Population Risk Rather Than True
Low Risk Without Any Risk Factors
In Figure 1 of the Science paper (1), lifetime cancer risk for
some tissues is plotted against estimated total stem cell divisions in that tissue on the log scale. In some cases, high- and
Received: June 12, 2015; Revised: August 10, 2015; Accepted: October 16, 2015
© The Author 2015. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected].
1 of 3
Figure 1. The effect of the same risk factor (adenomatous polyposis coli [APC]) gene mutation is supra-additive, resulting in many more colorectal cancers than duodenal
cancers in familial polyposis coli patients (1). If the amount of mutations from APC and from stem cell division were independent and additive, we would predict only
an 8.3% lifetime risk of colorectal cancer because of FAP, 4.5% because of baseline stem cell division, and 3.5% because of APC mutation, estimated from duodenal cancers. The observed excess of cases of colorectal cancer because of FAP is much higher than the observed excess cases of duodenal cancer because of FAP.
low-risk groups for the same cancer site were plotted (eg,
familial syndrome colorectal cancer, colorectal cancer; lung
cancer [smokers], lung cancer [nonsmokers]; HCV hepatocellular cancer, hepatocellular cancer). The “non-risk factor”
group was drawn from US national averages. It is important
to note that these groups, without presumed risk factors,
are not truly low-risk groups. For example, the rates of the
three cancers displayed with the highest rates but presumably without a risk factor (colorectal cancer [4.8% lifetime
risk], head and neck cancer [1.4%], and skin cancer [basal cell
{30%} and melanoma {2%}]) are based on the average lifetime
risk in the United States and actually incorporate the influence of risk factors. For example, non–human papillomavirus
(HPV) head and neck cancers are largely a result of tobacco
and alcohol (2). For colorectal cancer, the lifetime risk of 4.8%
includes a mixture of established or likely risk factors, including tobacco, alcohol, obesity, sedentary lifestyle, lack of physical activity, and various dietary exposures (3). The achievable
lifetime risk, by avoiding risk factors, is theoretically less than
1%, as evidenced by the relative rarity of colorectal cancer
before industrial times, in low-risk countries, and low rates
identified even in the US population among those who avoid
most risk factors (4). Basal cell and melanoma skin cancers,
which have relatively high population rates, are known to be
caused largely by solar ultraviolet (UV) exposure, with other
potential contributing factors such as HPV and arsenic. Rates
in nonwhites are greater than ten-fold lower than those in
whites, and most melanomas have mutational signatures of
UV damage (5), demonstrating that UV exposure likely causes
the majority of skin cancers.
The implication of the above is that even for tissues with
high stem cell division rates, it is unclear that lifetime risk
would exceed 1%, except possibly for basal cell skin cancer, a
largely nonfatal cancer. Thus, although the basic premise of the
paper that stem cell division correlates with tissue cancer rate
could be correct, it is also likely that lifetime cancer rate in an
organ in the absence of an additional risk factor may be quite
low, typically below 1% lifetime risk.
Stem Cell Division and Hereditary/
Environmental Risk Factors Are Likely to
Act Multiplicatively, Not Additively
A key aspect of the argument is summarized by authors: “The
potential causes of mutations, and therefore cancer, can be partitioned in two subsets: factors related to the number of stem
cell divisions and factors unrelated to those divisions. Thus, by
assumption, these two causes add up to explain exactly 100% of
the variation in risk” (6). Risk factors can act in a diverse number
of ways physically, but some, possibly most, risk factors multiply
or synergize the effect of baseline stem cell divisions rather than
simply add independently to the mutations. In fact, the authors of
the Science paper make precisely this point: “For example, patients
with FAP are ~30 times as likely to develop colorectal cancer than
duodenal cancer.... Our data suggest that this is because there
are ~150 times as many stem cell divisions in the colon as in the
duodenum. The lifetime risk of colorectal cancer would be very
low, even in the presence of an underlying APC gene mutation, if
colonic epithelial stem cells were not constantly dividing.”
The two quoted statements in the preceding paragraph offer
different perspectives; I believe the second statement is most
likely to be correct in general. To compare the implications of
these two statements, stem cell division in the colorectum and
duodenum in the absence or presence of an underlying APC
gene mutation is plotted against lifetime cancer risk. As seen in
Figure 1, the APC gene mutation is associated with approximately
30 times as many excess colorectal cancers as duodenal cancers.
If the APC mutation simply added excess cancers independently
of stem cell proliferation, the number of excess colorectal cancers
should approximately equal that of excess duodenal cancers.
The assumption of additivity, that two factors act completely
independently of each other, is required to be able to partition
causes of cancer into two subsets, which must add up to 100%.
However, it is unrealistic to consider stem cell divisions act completely independent of other risk factors. In fact, the pool of dividing
stem cells is what a risk factor is physically acting upon. The APC
example clearly shows this—the same mutation (risk factor) causes
commentary
E. L. Giovannucci | 2 of 3
3 of 3 | JNCI J Natl Cancer Inst, 2016, Vol. 108, No. 3
commentary
Figure 2. Figure demonstrating conceptually that the large (2200-fold) difference between stem cell proliferation could account for 150-fold differential in lung adenocarcinoma (nonsmokers) over pelvic osteosarcoma. Despite the 150-fold difference, the absolute lifetime risks are low (0.003% vs 0.45%). Although the excess risk of
lung adenocarcinoma because of smoking is relatively low (18-fold), it pushes lifetime risk to 8.1% and thus could account for the vast majority of lung cancers in a
population with high smoking rates.
many more cancers in a highly proliferative organ than in a lowly
proliferative one. We cannot examine a colorectal cancer in a FAP
patient and determine whether it was “caused” by the APC mutation or by the high proliferation rate in the colorectum; both factors
probably contributed. It is unlikely that most risk factors (eg, defect
in DNA repair, a chemical mutagen, inadequate supply of DNA
precursors required during replication, ionizing radiation, and hormones and growth factors, which may directly affect proliferation)
will increase mutations and ultimately cancers completely independent of the underlying proliferative component of the tissue.
That the Stem Cell Proliferation Rate Could
Contribute Largely to Differences in Cancer
Rates Across Tissues Does Not Preclude
Risk Factors From Contributing to the
Majority of Cancers in a Population
As aptly summarized by the authors (1): “One of the most impressive features of this correlation was that it extended across five
orders of magnitude, thereby applying to cancers with enormous
differences in incidence. No other environmental or inherited factors are known to be correlated in this way across tumor types.” For
example, relative to nonsmokers, the lung cancer rate in smokers
is about 20-fold higher, much less than five orders of magnitude.
The authors’ point that cell division rate variation across organs is
substantial and may underlie large differences in cancer susceptibilities of different organs may be valid, but if we accept the two
main conclusions above, it is still likely that most of the cancers
that arise in a population can be attributed to risk factors. To summarize the points above: 1) While stem cell division might be a
strong determinant of cancer risk in a tissue, without any additional risk factor the absolute lifetime risk of that cancer is low (eg,
<1%); and 2) risk factors are likely to multiply the baseline cancer
risk, which is influenced by stem cell division.
To see how both propositions 1) that stem cell division
explains most of the variation in differences of rates across tissues and 2) that risk factors can still account for most cancers
in a population can be valid, refer to Figure 2. A straight line is
drawn, representing the correlation between stem cell division
and lifetime cancer risk; on this line, the actual numbers for pelvic osteosarcoma and lung adenocarcinoma (nonsmokers) are
used to form the slope (1). Three data points are used to illustrate
three broad categories for cancers: low stem cell divisions (pelvic
osteosarcoma: lifetime risk: 0.003%), high stem cell division without risk factor (lung adenocarcinoma; nonsmokers: lifetime risk:
0.45%), and high stem cell division with risk factor (lung adenocarcinoma; smokers: lifetime risk: 8.1%). The same principle would
apply for multiple cancer types: Organs with very low stem proliferation will yield very low lifetime risk of cancer, organs with
relatively high stem proliferation will yield higher but still relative low lifetime risk, and organs with relatively high stem proliferation superimposed with an important risk factor will yield
high lifetime risk. Generalizing across the spectrum of organs, the
majority of cancers in absolute terms will tend to occur in organs
with high stem cell proliferation and with a strong risk factor, and
thus are potentially preventable.
The paper by Tomasetti and Vogelstein offers insights to
why cancer rates vary greatly across tissues. Their paper is an
important first step; enhancements to this model would involve
improving the estimates of lifetime organ stem cell proliferation
and correlating these rates to better estimates of lifetime cancer
risk in organs, removing the influence of risk factors. This would
provide a better estimate of the cancers that may be considered
to be attributable mostly to stem cell proliferation.
References
1. Tomasetti C, Vogelstein B. Cancer etiology. Variation in cancer risk among
tissues can be explained by the number of stem cell divisions. Science.
2015;347(6217):78–81.
2. Anantharaman D. Population attributable risk of tobacco and alcohol for
upper aerodigestive tract cancer. Oral Oncol. 2011;47(8):725–731.
3. Chan AT, Giovannucci EL. Primary prevention of colorectal cancer. Gastroenterology. 2010;138(6):2029–2043.
4. Platz EA, Willett WC, Colditz GA, Rimm EB, Spiegelman D, Giovannucci E. Proportion of colon cancer risk that might be preventable in a cohort of middleaged US men. Cancer Causes Control. 2000;11(7):579–588.
5. Mar VJ, Wong SQ, Li J, et al. BRAF/NRAS wild-type melanomas have a high
mutation load correlating with histologic and molecular signatures of UV
damage. Clin Cancer Res. 2013;19(17):4589–4598.
6. Tomasetti C, Vogelstein B. Cancer risk: role of environment—response. Science. 2015;347(6223):729–731.