more holistic approach to evaluation, respect for
consumer autonomy in decision making, and pursuit of
informed choice in screening should lead to better
benefit:harm ratios and improve the framework for
assessing the worth of population screening pro-
Contributors and sources: LI is a medical epidemiologist who
has researched issues in screening for several decades. KMcC is
a behavioural scientist with expertise in decision making and the
psychosocial outcomes of screening. GS has published widely
on consumer preferences for screening and has a long-standing
interest in the application of economic methods for the evalua-
tion of screening. PB is a clinical epidemiologist who studies
methods to evaluate medical tests and to obtain patient
trade-offs of benefit and harm. This article arose from concerns
about how the incentive to achieve high participation in screen-
ing often runs counter to principles of informed choice. All
authors contributed to the development of ideas and the writing
of the manuscript. LI is the guarantor.
Funding: Supported in part by Australian National Health and
Medical Research Council NHMRC grants 211205 and 402764
to the Screening and Test Evaluation Program. The NHMRC
had no role in the writing this article.
Competing interest: None declared.
1 Welch HG. Should I be tested for cancer? Maybe not and here’s why. Berkeley:
University of California Press, 2004.
2 Salkeld G. What are the benefits of preventive health care? Health Care
UK National Screening Committee. Criteria for appraising the viability,
effectiveness and appropriateness of a screening programme. www.nsc.nhs.uk/
pdfs/criteria.pdf (accessed 1 Feb 2006).
Strong K,Wald N,Miller A,Alwan A,WHO Consultation Group.Current
concepts in screening for noncommunicable disease: World Health
Organization consultation group report on methodology of noncommu-
nicable disease screening. J Med Screen 2005;12:12-9.
Barratt A,Howard K,Irwig L,Salkeld G,Houssami N.Model of outcomes
of screening mammography: information to support informed choices.
Cockburn J, Pit S, Redman S. Perceptions of screening mammography
among women aged 40-49. Aust N Z J Pub Health 1999;23:318-21.
Abelson J, Forest PG, Eyles J, Smith P, Martin E, Gauvin FP. Deliberations
about deliberative methods: issues in the design and evaluation of public
participation processes. Soc Sci Med 2003;57:239-51.
Irwig L, Glasziou P. Informed consent for screening by community
sampling. Effect Clin Pract 2000;2(3):47-50.
Salkeld G, Solomon M, Short L, Ryan M, Ward JE. Evidence-based
consumer choice: a case study in colorectal cancer screening. Aust N Z J
Pub Health 2003;27:449-55.
10 Barratt A, Trevena L, Davey HM, McCaffery K. Use of decision aids to
support informed choices about screening. BMJ 2003;329:507-10.
11 NHSCancer Screening Programme.
(accessed 1 Feb 2006).
12 NHS Cancer ScreeningProgramme.
1 Feb 2006).
13 Jorgensen KJ, Gotzsche PC. Presentation on websites of possible benefits
and harms from screening for breast cancer: cross section study. BMJ
14 Slaytor EK, Ward JE. How risks of breast cancer and benefits of screening
are communicated to women: analysis
15 Schwartz LM,Woloshin S,Fowler FJ Jr,Welch HG.Enthusiasm for cancer
screening in the United States. JAMA 2004;291:71-8.
16 O’Connor AM, Stacey D, Entwistle V, Llewellyn-Thomas H, Rovner D,
Holmes-Rovner M,et al.Decision aids for people facing health treatment
or screening decisions. Cochrane Database Syst Rev 2003;(2):CD001431.
17 Jepson RG, Hewison J, Thompson AGH, Weller D. How should we meas-
ure informed choice? The case of cancer screening. J Med Ethics
18 Raffle AE.Information about screening—is it to achieve high uptake or to
ensure informed choice? Health Expect 2001;4:92-8.
19 AustokerJ. Gaininginformed
20 Torgerson DJ, Donaldson C. An economic view of high compliance as a
screening objective. BMJ 1994;308:117-9.
21 Howard K, Salkeld G, Irwig L, Adelstein BA. High participation rates are
not necessary for cost-effective colorectal cancer screening. J Med Screen
22 Marteau TM, Dormandy E, Michie S. A measure of informed choice.
Health Expect 2001;4:99-108.
23 Dowie J. The role of patients’ meta preferences in the design and evalua-
tion of decision support systems. Health Expect 2002;5:16-27.
Breast screening:the facts.
Cervical cancer:the facts.
of 58pamphlets. BMJ
(Accepted 11 March 2006)
Common susceptibility genes for cancer: search for the
end of the rainbow
Stuart G Baker, Jaakko Kaprio
The human genome map has started a hunt to find common genes that are associated with cancer.
But new research questions the likelihood of success.
Huge resources are being invested in the search for
common inherited genetic variants that increase
susceptibility to cancer. However, these studies are
expensive because they require large sample sizes to
rule out false positive results (table).1 2The US cancer
genetic markers of susceptibility project (http://
cgems.cancer.gov), for example, will cost $14m (£7.9m;
€11m).In addition,large replication studies may still be
necessary to confirm generalisability to other popula-
tions. For these studies to eventually lead to a clinical
therapeutic benefit, common genetic variants that
increase susceptibility to cancer must exist and it must
be feasible to rigorously evaluate the clinical benefit of
targeting these common genetic variants. Both these
requirements require formal consideration.
All benefits and harms of screening must be fully
ascertained and undergo community valuation
before screening is offered
Service providers should respect patient
autonomy and ensure that participation in
screening is an informed choice
Concordance between consumer preferences and
screening behaviour should replace participation
as one of the measures of success for screening
An explanation of sample size calculations for gene-cancer
association studies is on bmj.com
Analysis and comment
Stuart G Baker
S G Baker
BMJ VOLUME 332 13 MAY 2006bmj.com
Common cancer susceptibility genes are
Devoting a large research effort to searching for com-
mon cancer susceptibility genes has several problems.
The first is that recent research suggests these genes
are unlikely to exist or, if they do, are unlikely to have
much of an effect on the incidence of cancer. The early
phases of carcinogenesis seem to entail alterations in
the stroma (supporting tissue) rather than a genetic
mutation of the parenchyma (functional tissue).3 4Thus
genetic susceptibility to cancer of the parenchyma
(except for rare genes related to familial cancer) would
have a relatively small role in the early stages of
carcinogenesis.Moreover,a recent study could not find
conclusive evidence of genetic alterations in the
stroma,5further diminishing the probable role of
genetic susceptibility in early stage carcinogenesis.
In addition, the rapid changes in cell morphology
needed for evolving cancer cells to have a growth
advantage over other cells are likely to require large
genetic rearrangements6rather than single polymor-
phic changes. Of course, genetic mutations could also
affect concentrations of hormones or growth factors,
which might affect the tumour microenvironment. But
these concentrations would also be affected by
A second reason to play down the role of common
genetic susceptibility genes
suggesting that environmental, dietary, or lifestyle
changes have a large effect on the incidence of
cancer.7 8These studies show changes in incidence
within one or two generations, which is probably too
quick to be related to the introduction of new cancer
A final reason to be sceptical of the role of
common genetic susceptibility mutations in the cause
of cancer comes from results from a Nordic study of
cancers in twins.9By analysing data from monozygotic
(identical) and dizygotic (fraternal) twins, the authors
showed that genetic susceptibility made only a small to
moderate contribution to the incidence of cancer.9The
results support the argument for the primacy of
environmental effects. Risch questioned the study’s
conclusions, using a mendelian model to show that the
data could be consistent with a gene for cancer suscep-
tibility with a low genetic relative risk.10However, the
model was unrealistic because it assumed independ-
ence of cancer incidence among twins without a cancer
susceptibility gene.11This is unlikely given that twins
tend to have shared experiences,exposures,habits,and
cancer screening behaviour (as would family members
A more realistic mendelian model that allowed
dependence of cancer incidence among twins without
genetic susceptibility estimated that the fraction of can-
cers with a susceptibility genotype would be 0.09 to
0.22 for prostate cancer, 0.08 to 0.14 for breast cancer,
and 0.05 to 0.13 for colorectal cancer.11Even these esti-
mates are probably high because environmental
factors were not modelled in the twin study and
environmental factors among twins are likely to be
similar. Moreover, the model fit to the twin data
estimated a low prevalence for a cancer susceptibility
genotype and a high genetic relative risk, as would be
found in rare genes typically identified by mendelian
inheritance patterns in pedigrees.12 13By contrast, most
research is focused on identifying common single
nucleotide polymorphisms with a low genetic relative
Studies purporting to show a high likelihood of
association between common low penetrance genes
and cancer rely on less plausible assumptions than the
Nordic twin study.For example,variations in the risk of
secondary cancers have been argued to be best
explained by variations in genetic risk.14–16However,the
variation in risk could also be explained by non-genetic
risk factors such as defects in cellular communication3 4
or epigenetic mechanisms such as methylation of
DNA. A meta-analysis of replicate studies after an
initial positive finding found an association between a
common gene and head and neck cancer.17Although
the authors discounted chance or bias in publishing
studies that had significant results, substantial residual
publication bias could have remained because of pref-
erential submission and publication of studies that do
not contradict the original reports even if the negative
results were significant. Importantly, the same study
found no association between a common genetic vari-
ant and breast cancer.
Difficulty showing clinical benefit
Even if we are incorrect, and if large studies detect true
associations between common genes and common
Sample sizes required to make false positive results unlikely
when testing association between genetic variant and cancer*
Genetic relative risk† Prior probability of association
Sample size for
*Based on a two sided type I error of 0.05, a power of 0.90, and a false
positive report probability of 0.05 (see bmj.com for calculation details)
†Relative risk of cancer in people with genetic variant compared with those
‡ An equal number of controls is also needed.
DNA chip analysis comparing gene expression in normal and
cancerous prostate cells
VERONIQUE BLANC AND QIN WANG/WELLCOME PHOTO LIBRARY
Analysis and comment
BMJ VOLUME 332 13 MAY 2006bmj.com
cancers, showing clinical benefit will still be difficult.
Ideally, identification of a gene related to cancer would
provide important biological insights that suggest a
modifiable risk factor or lead to a new treatment. The
first step to prove clinical benefit would be additional
observational studies to confirm the effect of the modi-
fiable risk factor. The second step would be a
randomised trial to determine if benefits outweigh
harms in asymptomatic people.Volunteers would need
to be tested for the genetic variant, and those who had
the variant randomised to an intervention or placebo.
Such trials are likely to be both expensive and long.
Because cancer is rare in asymptomatic people, very
large sample sizes are required, typically 10 000 to
30 000.18Even larger numbers of people have to
have genetic testing to identify those suitable for
If a randomised trial has already been conducted of
an intervention targeted at the likely modifiable risk
factor, an alternative design can be used that reduces
the amount of genetic testing. The idea is to “piggy
back” on the existing trial by using a nested
case-control design to select all participants with
cancer and a random sample without cancer for
genetic testing. Such a nested case-control study was
proposed for the further analysis of data from the
breast cancer prevention trial, which showed that
tamoxifen reduced the risk of breast cancer but at the
substantial cost of harmful side effects.19The goal of
the nested case-control study was to identify a subset of
participants with a gene who might have a greater
reduction in breast cancer risk than the average
participant so that benefits would outweigh harms. In
the design calculations two types of genes were consid-
ered: rare genes with high genetic relative risk and
common genes with low relative risk. It was found that
only rare genes with high relative risk would be likely to
be detected with a benefit that outweighed harms. If
this case is a guide, the approach would not be suitable
for use with common genes.
The search for common cancer susceptibility genes
faces important methodological and practical chal-
lenges for cancer prevention, given the small chance
that such genetic variants exist and the difficulty and
expense of proving substantial clinical benefit if they
do exist. Enthusiasm for this new field of research
should not precipitate unwarranted expectations.
Contributors and sources: SGB and JK have analysed Nordic
twin data and SGB has investigated nested case-control studies
for studying genetic effects in randomised trials. JK works as a
genetic epidemiologist for the Finnish National Public Health
Institute.SGB wrote the initial draft and JK provided substantive
comments and substantial editing. SGB is the guarantor.
Funding: JK is supported by the Genomeutwin project
(European Union contract QLG2-CT-2002-01254).
Competing interests: None declared.
1Colhoun HM, McKeigue PM, Smith DS. Problems of reporting genetic
associations with complex outcomes. Lancet 2003;361:865-72.
Ioannidis JPA, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG.
Replication validityof geneticassociation
Soto AM, Sonnenschein C. The somatic mutation theory of cancer:
growing problems with the paradigm? BioEssays 2004;26:1097-107.
4 Tisty TD, Hein PW. Know thy neighbor: stromal cells can contribute
oncogenic signals. Curr Opin Genet Dev 2001;11:54-9.
Allinen M, Beroukhim R, Cai L, Brennan C, Lahti-Domenici J, Huang H,
et al. Molecular characterization of the tumor microenvironment in
breast cancer. Cancer Cell 2004;6:17-32.
Cairns J. The origin of human cancers. Nature 1981;289:353-67.
Ziegler RG,Hoover RN,Pike MC,Hildesheim A,Nomura AMY,West DW,
et al. Migration patterns and breast cancer risk in Asian-American
women. J Natl Cancer Inst 1993;85:1819-27.
Tominaga S. Cancer incidence in Japanese in Japan, Hawaii, and Western
United States. Natl Cancer Inst Monogr 1985;69:83-92.
Lichtenstein P, Holm NV, Verkasalo PK, Iliadou A, Kaprio J, Koskenvuo
M, et al. Environmental and heritable factors in the causation of cancer.
Analyses of cohorts of twins from Sweden,Denmark,and Finland.N Engl
J Med 2000;343:78-85.
10 Risch N. The genetic epidemiology of cancer: interpreting family and
twin studies and their implications for molecular genetic approaches.
Cancer Epidemiol Biomarkers Prev 2001;10:733-41.
11 Baker SG, Lichtenstein P, Kaprio J, Holms N. Genetic susceptibility to
prostate, breast and colorectal cancer among Nordic twins. Biometrics
12 Calvert PM, Frucht H. The genetics of colorectal cancer. Ann Intern Med
13 Hopper JL. More breast cancer genes? Breast Cancer Res 2001;3:154-7.
14 Begg CB. The search for cancer risk factors: when can we stop looking?
Am J Public Health 2001;91:360-4.
15 Berwick M, Begg CB, Fine JA, Roush GC, Barnhill RL. Screening for
cutaneous melanoma by skin self-examination. J Natl Cancer Inst
16 Begg CB, Orlow I, Hummer AJ, Armstrong BK, Kricker A, Marrett LD, et
al. Lifetime risk of melanoma in CDKN2A mutation carriers in a
population-based sample. J Natl Cancer Inst 2005;98:1507-15.
17 Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN.
Meta-analysis of genetic association studies supports a contribution of
common variants to susceptibility to common disease. Nature Genet
18 Tsao AS, Kim ES, Hong WK. Chemoprevention of cancer. CA Cancer J
19 Baker SG, Kramer BS. Statistics for weighing benefits and harms in a
proposed genetic substudy of a randomized cancer prevention trial. J R
Stat Soc Ser C Appl Stat 2005;54:941-54.
(Accepted 9 March 2006)
Studies to detect common cancer susceptibility
genes are expensive because they require large
Evidence from biology, migration studies, and
twin studies suggests that common cancer
susceptibility genes are unlikely
Even if susceptibility genes were identified, further
large studies would be needed to show clinical
Making themselves sicke by
Dr Wedderburne gave mee this Medicall
Counsayle. That I should not be sicke before I was
sicke, Noteing the puling spirits of some, &
generally of Scholars, who doe allwayes phantsy
themselves to be sicke, or sickly, & by phantsye &
frequent medicines doe make themselves sick
John Beale writing to Robert Boyle in 1663.
Quoted in The curious life of Robert Hooke by
Lisa Jardine, Harper Perennial, London, 2003:215.)
Elizabeth Wager, publications consultant, Princes
Analysis and comment
BMJ VOLUME 33213 MAY 2006bmj.com